Merge pull request #832 from mvanhorn/feat/evals-promptfoo-bootstrap

feat(evals): bootstrap promptfoo eval framework (Phase 0)

.agents/skills/company-creator/SKILL.md

@@ -0,0 +1,269 @@
+---
+name: company-creator
+description: >
+  Create agent company packages conforming to the Agent Companies specification
+  (agentcompanies/v1). Use when a user wants to create a new agent company from
+  scratch, build a company around an existing git repo or skills collection, or
+  scaffold a team/department of agents. Triggers on: "create a company", "make me
+  a company", "build a company from this repo", "set up an agent company",
+  "create a team of agents", "hire some agents", or when given a repo URL and
+  asked to turn it into a company. Do NOT use for importing an existing company
+  package (use the CLI import command instead) or for modifying a company that
+  is already running in Paperclip.
+---
+
+# Company Creator
+
+Create agent company packages that conform to the Agent Companies specification.
+
+Spec references:
+
+- Normative spec: `docs/companies/companies-spec.md` (read this before generating files)
+- Web spec: https://agentcompanies.io/specification
+- Protocol site: https://agentcompanies.io/
+
+## Two Modes
+
+### Mode 1: Company From Scratch
+
+The user describes what they want. Interview them to flesh out the vision, then generate the package.
+
+### Mode 2: Company From a Repo
+
+The user provides a git repo URL, local path, or tweet. Analyze the repo, then create a company that wraps it.
+
+See [references/from-repo-guide.md](references/from-repo-guide.md) for detailed repo analysis steps.
+
+## Process
+
+### Step 1: Gather Context
+
+Determine which mode applies:
+
+- **From scratch**: What kind of company or team? What domain? What should the agents do?
+- **From repo**: Clone/read the repo. Scan for existing skills, agent configs, README, source structure.
+
+### Step 2: Interview (Use AskUserQuestion)
+
+Do not skip this step. Use AskUserQuestion to align with the user before writing any files.
+
+**For from-scratch companies**, ask about:
+
+- Company purpose and domain (1-2 sentences is fine)
+- What agents they need - propose a hiring plan based on what they described
+- Whether this is a full company (needs a CEO) or a team/department (no CEO required)
+- Any specific skills the agents should have
+- How work flows through the organization (see "Workflow" below)
+- Whether they want projects and starter tasks
+
+**For from-repo companies**, present your analysis and ask:
+
+- Confirm the agents you plan to create and their roles
+- Whether to reference or vendor any discovered skills (default: reference)
+- Any additional agents or skills beyond what the repo provides
+- Company name and any customization
+- Confirm the workflow you inferred from the repo (see "Workflow" below)
+
+**Workflow — how does work move through this company?**
+
+A company is not just a list of agents with skills. It's an organization that takes ideas and turns them into work products. You need to understand the workflow so each agent knows:
+
+- Who gives them work and in what form (a task, a branch, a question, a review request)
+- What they do with it
+- Who they hand off to when they're done, and what that handoff looks like
+- What "done" means for their role
+
+**Not every company is a pipeline.** Infer the right workflow pattern from context:
+
+- **Pipeline** — sequential stages, each agent hands off to the next. Use when the repo/domain has a clear linear process (e.g. plan → build → review → ship → QA, or content ideation → draft → edit → publish).
+- **Hub-and-spoke** — a manager delegates to specialists who report back independently. Use when agents do different kinds of work that don't feed into each other (e.g. a CEO who dispatches to a researcher, a marketer, and an analyst).
+- **Collaborative** — agents work together on the same things as peers. Use for small teams where everyone contributes to the same output (e.g. a design studio, a brainstorming team).
+- **On-demand** — agents are summoned as needed with no fixed flow. Use when agents are more like a toolbox of specialists the user calls directly.
+
+For from-scratch companies, propose a workflow pattern based on what they described and ask if it fits.
+
+For from-repo companies, infer the pattern from the repo's structure. If skills have a clear sequential dependency (like `plan-ceo-review → plan-eng-review → review → ship → qa`), that's a pipeline. If skills are independent capabilities, it's more likely hub-and-spoke or on-demand. State your inference in the interview so the user can confirm or adjust.
+
+**Key interviewing principles:**
+
+- Propose a concrete hiring plan. Don't ask open-ended "what agents do you want?" - suggest specific agents based on context and let the user adjust.
+- Keep it lean. Most users are new to agent companies. A few agents (3-5) is typical for a startup. Don't suggest 10+ agents unless the scope demands it.
+- From-scratch companies should start with a CEO who manages everyone. Teams/departments don't need one.
+- Ask 2-3 focused questions per round, not 10.
+
+### Step 3: Read the Spec
+
+Before generating any files, read the normative spec:
+
+```
+docs/companies/companies-spec.md
+```
+
+Also read the quick reference: [references/companies-spec.md](references/companies-spec.md)
+
+And the example: [references/example-company.md](references/example-company.md)
+
+### Step 4: Generate the Package
+
+Create the directory structure and all files. Follow the spec's conventions exactly.
+
+**Directory structure:**
+
+```
+<company-slug>/
+├── COMPANY.md
+├── agents/
+│   └── <slug>/AGENTS.md
+├── teams/
+│   └── <slug>/TEAM.md        (if teams are needed)
+├── projects/
+│   └── <slug>/PROJECT.md     (if projects are needed)
+├── tasks/
+│   └── <slug>/TASK.md        (if tasks are needed)
+├── skills/
+│   └── <slug>/SKILL.md       (if custom skills are needed)
+└── .paperclip.yaml            (Paperclip vendor extension)
+```
+
+**Rules:**
+
+- Slugs must be URL-safe, lowercase, hyphenated
+- COMPANY.md gets `schema: agentcompanies/v1` - other files inherit it
+- Agent instructions go in the AGENTS.md body, not in .paperclip.yaml
+- Skills referenced by shortname in AGENTS.md resolve to `skills/<shortname>/SKILL.md`
+- For external skills, use `sources` with `usage: referenced` (see spec section 12)
+- Do not export secrets, machine-local paths, or database IDs
+- Omit empty/default fields
+- For companies generated from a repo, add a references footer at the bottom of COMPANY.md body:
+  `Generated from [repo-name](repo-url) with the company-creator skill from [Paperclip](https://github.com/paperclipai/paperclip)`
+
+**Reporting structure:**
+
+- Every agent except the CEO should have `reportsTo` set to their manager's slug
+- The CEO has `reportsTo: null`
+- For teams without a CEO, the top-level agent has `reportsTo: null`
+
+**Writing workflow-aware agent instructions:**
+
+Each AGENTS.md body should include not just what the agent does, but how they fit into the organization's workflow. Include:
+
+1. **Where work comes from** — "You receive feature ideas from the user" or "You pick up tasks assigned to you by the CTO"
+2. **What you produce** — "You produce a technical plan with architecture diagrams" or "You produce a reviewed, approved branch ready for shipping"
+3. **Who you hand off to** — "When your plan is locked, hand off to the Staff Engineer for implementation" or "When review passes, hand off to the Release Engineer to ship"
+4. **What triggers you** — "You are activated when a new feature idea needs product-level thinking" or "You are activated when a branch is ready for pre-landing review"
+
+This turns a collection of agents into an organization that actually works together. Without workflow context, agents operate in isolation — they do their job but don't know what happens before or after them.
+
+### Step 5: Confirm Output Location
+
+Ask the user where to write the package. Common options:
+
+- A subdirectory in the current repo
+- A new directory the user specifies
+- The current directory (if it's empty or they confirm)
+
+### Step 6: Write README.md and LICENSE
+
+**README.md** — every company package gets a README. It should be a nice, readable introduction that someone browsing GitHub would appreciate. Include:
+
+- Company name and what it does
+- The workflow / how the company operates
+- Org chart as a markdown list or table showing agents, titles, reporting structure, and skills
+- Brief description of each agent's role
+- Citations and references: link to the source repo (if from-repo), link to the Agent Companies spec (https://agentcompanies.io/specification), and link to Paperclip (https://github.com/paperclipai/paperclip)
+- A "Getting Started" section explaining how to import: `paperclipai company import --from <path>`
+
+**LICENSE** — include a LICENSE file. The copyright holder is the user creating the company, not the upstream repo author (they made the skills, the user is making the company). Use the same license type as the source repo (if from-repo) or ask the user (if from-scratch). Default to MIT if unclear.
+
+### Step 7: Write Files and Summarize
+
+Write all files, then give a brief summary:
+
+- Company name and what it does
+- Agent roster with roles and reporting structure
+- Skills (custom + referenced)
+- Projects and tasks if any
+- The output path
+
+## .paperclip.yaml Guidelines
+
+The `.paperclip.yaml` file is the Paperclip vendor extension. It configures adapters and env inputs per agent.
+
+### Adapter Rules
+
+**Do not specify an adapter unless the repo or user context warrants it.** If you don't know what adapter the user wants, omit the adapter block entirely — Paperclip will use its default. Specifying an unknown adapter type causes an import error.
+
+Paperclip's supported adapter types (these are the ONLY valid values):
+- `claude_local` — Claude Code CLI
+- `codex_local` — Codex CLI
+- `opencode_local` — OpenCode CLI
+- `pi_local` — Pi CLI
+- `cursor` — Cursor
+- `gemini_local` — Gemini CLI
+- `openclaw_gateway` — OpenClaw gateway
+
+Only set an adapter when:
+- The repo or its skills clearly target a specific runtime (e.g. gstack is built for Claude Code, so `claude_local` is appropriate)
+- The user explicitly requests a specific adapter
+- The agent's role requires a specific runtime capability
+
+### Env Inputs Rules
+
+**Do not add boilerplate env variables.** Only add env inputs that the agent actually needs based on its skills or role:
+- `GH_TOKEN` for agents that push code, create PRs, or interact with GitHub
+- API keys only when a skill explicitly requires them
+- Never set `ANTHROPIC_API_KEY` as a default empty env variable — the runtime handles this
+
+Example with adapter (only when warranted):
+```yaml
+schema: paperclip/v1
+agents:
+  release-engineer:
+    adapter:
+      type: claude_local
+      config:
+        model: claude-sonnet-4-6
+    inputs:
+      env:
+        GH_TOKEN:
+          kind: secret
+          requirement: optional
+```
+
+Example — only agents with actual overrides appear:
+```yaml
+schema: paperclip/v1
+agents:
+  release-engineer:
+    inputs:
+      env:
+        GH_TOKEN:
+          kind: secret
+          requirement: optional
+```
+
+In this example, only `release-engineer` appears because it needs `GH_TOKEN`. The other agents (ceo, cto, etc.) have no overrides, so they are omitted entirely from `.paperclip.yaml`.
+
+## External Skill References
+
+When referencing skills from a GitHub repo, always use the references pattern:
+
+```yaml
+metadata:
+  sources:
+    - kind: github-file
+      repo: owner/repo
+      path: path/to/SKILL.md
+      commit: <full SHA from git ls-remote or the repo>
+      attribution: Owner or Org Name
+      license: <from the repo's LICENSE>
+      usage: referenced
+```
+
+Get the commit SHA with:
+
+```bash
+git ls-remote https://github.com/owner/repo HEAD
+```
+
+Do NOT copy external skill content into the package unless the user explicitly asks.

.agents/skills/company-creator/references/companies-spec.md

@@ -0,0 +1,144 @@
+# Agent Companies Specification Reference
+
+The normative specification lives at:
+
+- Web: https://agentcompanies.io/specification
+- Local: docs/companies/companies-spec.md
+
+Read the local spec file before generating any package files. The spec defines the canonical format and all frontmatter fields. Below is a quick-reference summary for common authoring tasks.
+
+## Package Kinds
+
+| File       | Kind    | Purpose                                           |
+| ---------- | ------- | ------------------------------------------------- |
+| COMPANY.md | company | Root entrypoint, org boundary and defaults        |
+| TEAM.md    | team    | Reusable org subtree                              |
+| AGENTS.md  | agent   | One role, instructions, and attached skills       |
+| PROJECT.md | project | Planned work grouping                             |
+| TASK.md    | task    | Portable starter task                             |
+| SKILL.md   | skill   | Agent Skills capability package (do not redefine) |
+
+## Directory Layout
+
+```
+company-package/
+├── COMPANY.md
+├── agents/
+│   └── <slug>/AGENTS.md
+├── teams/
+│   └── <slug>/TEAM.md
+├── projects/
+│   └── <slug>/
+│       ├── PROJECT.md
+│       └── tasks/
+│           └── <slug>/TASK.md
+├── tasks/
+│   └── <slug>/TASK.md
+├── skills/
+│   └── <slug>/SKILL.md
+├── assets/
+├── scripts/
+├── references/
+└── .paperclip.yaml          (optional vendor extension)
+```
+
+## Common Frontmatter Fields
+
+```yaml
+schema: agentcompanies/v1
+kind: company | team | agent | project | task
+slug: url-safe-stable-identity
+name: Human Readable Name
+description: Short description for discovery
+version: 0.1.0
+license: MIT
+authors:
+  - name: Jane Doe
+tags: []
+metadata: {}
+sources: []
+```
+
+- `schema` usually appears only at package root
+- `kind` is optional when filename makes it obvious
+- `slug` must be URL-safe and stable
+- exporters should omit empty or default-valued fields
+
+## COMPANY.md Required Fields
+
+```yaml
+name: Company Name
+description: What this company does
+slug: company-slug
+schema: agentcompanies/v1
+```
+
+Optional: `version`, `license`, `authors`, `goals`, `includes`, `requirements.secrets`
+
+## AGENTS.md Key Fields
+
+```yaml
+name: Agent Name
+title: Role Title
+reportsTo: <agent-slug or null>
+skills:
+  - skill-shortname
+```
+
+- Body content is the agent's default instructions
+- Skills resolve by shortname: `skills/<shortname>/SKILL.md`
+- Do not export machine-specific paths or secrets
+
+## TEAM.md Key Fields
+
+```yaml
+name: Team Name
+description: What this team does
+slug: team-slug
+manager: ../agent-slug/AGENTS.md
+includes:
+  - ../agent-slug/AGENTS.md
+  - ../../skills/skill-slug/SKILL.md
+```
+
+## PROJECT.md Key Fields
+
+```yaml
+name: Project Name
+description: What this project delivers
+owner: agent-slug
+```
+
+## TASK.md Key Fields
+
+```yaml
+name: Task Name
+assignee: agent-slug
+project: project-slug
+schedule:
+  timezone: America/Chicago
+  startsAt: 2026-03-16T09:00:00-05:00
+  recurrence:
+    frequency: weekly
+    interval: 1
+    weekdays: [monday]
+    time: { hour: 9, minute: 0 }
+```
+
+## Source References (for external skills/content)
+
+```yaml
+sources:
+  - kind: github-file
+    repo: owner/repo
+    path: path/to/SKILL.md
+    commit: <full-sha>
+    sha256: <hash>
+    attribution: Owner Name
+    license: MIT
+    usage: referenced
+```
+
+Usage modes: `vendored` (bytes included), `referenced` (pointer only), `mirrored` (cached locally)
+
+Default to `referenced` for third-party content.

.agents/skills/company-creator/references/example-company.md

@@ -0,0 +1,184 @@
+# Example Company Package
+
+A minimal but complete example of an agent company package.
+
+## Directory Structure
+
+```
+lean-dev-shop/
+├── COMPANY.md
+├── agents/
+│   ├── ceo/AGENTS.md
+│   ├── cto/AGENTS.md
+│   └── engineer/AGENTS.md
+├── teams/
+│   └── engineering/TEAM.md
+├── projects/
+│   └── q2-launch/
+│       ├── PROJECT.md
+│       └── tasks/
+│           └── monday-review/TASK.md
+├── tasks/
+│   └── weekly-standup/TASK.md
+├── skills/
+│   └── code-review/SKILL.md
+└── .paperclip.yaml
+```
+
+## COMPANY.md
+
+```markdown
+---
+name: Lean Dev Shop
+description: Small engineering-focused AI company that builds and ships software products
+slug: lean-dev-shop
+schema: agentcompanies/v1
+version: 1.0.0
+license: MIT
+authors:
+  - name: Example Org
+goals:
+  - Build and ship software products
+  - Maintain high code quality
+---
+
+Lean Dev Shop is a small, focused engineering company. The CEO oversees strategy and coordinates work. The CTO leads the engineering team. Engineers build and ship code.
+```
+
+## agents/ceo/AGENTS.md
+
+```markdown
+---
+name: CEO
+title: Chief Executive Officer
+reportsTo: null
+skills:
+  - paperclip
+---
+
+You are the CEO of Lean Dev Shop. You oversee company strategy, coordinate work across the team, and ensure projects ship on time.
+
+Your responsibilities:
+
+- Review and prioritize work across projects
+- Coordinate with the CTO on technical decisions
+- Ensure the company goals are being met
+```
+
+## agents/cto/AGENTS.md
+
+```markdown
+---
+name: CTO
+title: Chief Technology Officer
+reportsTo: ceo
+skills:
+  - code-review
+  - paperclip
+---
+
+You are the CTO of Lean Dev Shop. You lead the engineering team and make technical decisions.
+
+Your responsibilities:
+
+- Set technical direction and architecture
+- Review code and ensure quality standards
+- Mentor engineers and unblock technical challenges
+```
+
+## agents/engineer/AGENTS.md
+
+```markdown
+---
+name: Engineer
+title: Software Engineer
+reportsTo: cto
+skills:
+  - code-review
+  - paperclip
+---
+
+You are a software engineer at Lean Dev Shop. You write code, fix bugs, and ship features.
+
+Your responsibilities:
+
+- Implement features and fix bugs
+- Write tests and documentation
+- Participate in code reviews
+```
+
+## teams/engineering/TEAM.md
+
+```markdown
+---
+name: Engineering
+description: Product and platform engineering team
+slug: engineering
+schema: agentcompanies/v1
+manager: ../../agents/cto/AGENTS.md
+includes:
+  - ../../agents/engineer/AGENTS.md
+  - ../../skills/code-review/SKILL.md
+tags:
+  - engineering
+---
+
+The engineering team builds and maintains all software products.
+```
+
+## projects/q2-launch/PROJECT.md
+
+```markdown
+---
+name: Q2 Launch
+description: Ship the Q2 product launch
+slug: q2-launch
+owner: cto
+---
+
+Deliver all features planned for the Q2 launch, including the new dashboard and API improvements.
+```
+
+## projects/q2-launch/tasks/monday-review/TASK.md
+
+```markdown
+---
+name: Monday Review
+assignee: ceo
+project: q2-launch
+schedule:
+  timezone: America/Chicago
+  startsAt: 2026-03-16T09:00:00-05:00
+  recurrence:
+    frequency: weekly
+    interval: 1
+    weekdays:
+      - monday
+    time:
+      hour: 9
+      minute: 0
+---
+
+Review the status of Q2 Launch project. Check progress on all open tasks, identify blockers, and update priorities for the week.
+```
+
+## skills/code-review/SKILL.md (with external reference)
+
+```markdown
+---
+name: code-review
+description: Thorough code review skill for pull requests and diffs
+metadata:
+  sources:
+    - kind: github-file
+      repo: anthropics/claude-code
+      path: skills/code-review/SKILL.md
+      commit: abc123def456
+      sha256: 3b7e...9a
+      attribution: Anthropic
+      license: MIT
+      usage: referenced
+---
+
+Review code changes for correctness, style, and potential issues.
+```

.agents/skills/company-creator/references/from-repo-guide.md

@@ -0,0 +1,79 @@
+# Creating a Company From an Existing Repository
+
+When a user provides a git repo (URL, local path, or tweet linking to a repo), analyze it and create a company package that wraps its content.
+
+## Analysis Steps
+
+1. **Clone or read the repo** - Use `git clone` for URLs, read directly for local paths
+2. **Scan for existing agent/skill files** - Look for SKILL.md, AGENTS.md, CLAUDE.md, .claude/ directories, or similar agent configuration
+3. **Understand the repo's purpose** - Read README, package.json, main source files to understand what the project does
+4. **Identify natural agent roles** - Based on the repo's structure and purpose, determine what agents would be useful
+
+## Handling Existing Skills
+
+Many repos already contain skills (SKILL.md files). When you find them:
+
+**Default behavior: use references, not copies.**
+
+Instead of copying skill content into your company package, create a source reference:
+
+```yaml
+metadata:
+  sources:
+    - kind: github-file
+      repo: owner/repo
+      path: path/to/SKILL.md
+      commit: <get the current HEAD commit SHA>
+      attribution: <repo owner or org name>
+      license: <from repo's LICENSE file>
+      usage: referenced
+```
+
+To get the commit SHA:
+```bash
+git ls-remote https://github.com/owner/repo HEAD
+```
+
+Only vendor (copy) skills when:
+- The user explicitly asks to copy them
+- The skill is very small and tightly coupled to the company
+- The source repo is private or may become unavailable
+
+## Handling Existing Agent Configurations
+
+If the repo has agent configs (CLAUDE.md, .claude/ directories, codex configs, etc.):
+- Use them as inspiration for AGENTS.md instructions
+- Don't copy them verbatim - adapt them to the Agent Companies format
+- Preserve the intent and key instructions
+
+## Repo-Only Skills (No Agents)
+
+When a repo contains only skills and no agents:
+- Create agents that would naturally use those skills
+- The agents should be minimal - just enough to give the skills a runtime context
+- A single agent may use multiple skills from the repo
+- Name agents based on the domain the skills cover
+
+Example: A repo with `code-review`, `testing`, and `deployment` skills might become:
+- A "Lead Engineer" agent with all three skills
+- Or separate "Reviewer", "QA Engineer", and "DevOps" agents if the skills are distinct enough
+
+## Common Repo Patterns
+
+### Developer Tools / CLI repos
+- Create agents for the tool's primary use cases
+- Reference any existing skills
+- Add a project maintainer or lead agent
+
+### Library / Framework repos
+- Create agents for development, testing, documentation
+- Skills from the repo become agent capabilities
+
+### Full Application repos
+- Map to departments: engineering, product, QA
+- Create a lean team structure appropriate to the project size
+
+### Skills Collection repos (e.g. skills.sh repos)
+- Each skill or skill group gets an agent
+- Create a lightweight company or team wrapper
+- Keep the agent count proportional to the skill diversity

.agents/skills/doc-maintenance/SKILL.md

@@ -0,0 +1,201 @@
+---
+name: doc-maintenance
+description: >
+  Audit top-level documentation (README, SPEC, PRODUCT) against recent git
+  history to find drift — shipped features missing from docs or features
+  listed as upcoming that already landed. Proposes minimal edits, creates
+  a branch, and opens a PR. Use when asked to review docs for accuracy,
+  after major feature merges, or on a periodic schedule.
+---
+
+# Doc Maintenance Skill
+
+Detect documentation drift and fix it via PR — no rewrites, no churn.
+
+## When to Use
+
+- Periodic doc review (e.g. weekly or after releases)
+- After major feature merges
+- When asked "are our docs up to date?"
+- When asked to audit README / SPEC / PRODUCT accuracy
+
+## Target Documents
+
+| Document | Path | What matters |
+|----------|------|-------------|
+| README | `README.md` | Features table, roadmap, quickstart, "what is" accuracy, "works with" table |
+| SPEC | `doc/SPEC.md` | No false "not supported" claims, major model/schema accuracy |
+| PRODUCT | `doc/PRODUCT.md` | Core concepts, feature list, principles accuracy |
+
+Out of scope: DEVELOPING.md, DATABASE.md, CLI.md, doc/plans/, skill files,
+release notes. These are dev-facing or ephemeral — lower risk of user-facing
+confusion.
+
+## Workflow
+
+### Step 1 — Detect what changed
+
+Find the last review cursor:
+
+```bash
+# Read the last-reviewed commit SHA
+CURSOR_FILE=".doc-review-cursor"
+if [ -f "$CURSOR_FILE" ]; then
+  LAST_SHA=$(cat "$CURSOR_FILE" | head -1)
+else
+  # First run: look back 60 days
+  LAST_SHA=$(git log --format="%H" --after="60 days ago" --reverse | head -1)
+fi
+```
+
+Then gather commits since the cursor:
+
+```bash
+git log "$LAST_SHA"..HEAD --oneline --no-merges
+```
+
+### Step 2 — Classify changes
+
+Scan commit messages and changed files. Categorize into:
+
+- **Feature** — new capabilities (keywords: `feat`, `add`, `implement`, `support`)
+- **Breaking** — removed/renamed things (keywords: `remove`, `breaking`, `drop`, `rename`)
+- **Structural** — new directories, config changes, new adapters, new CLI commands
+
+**Ignore:** refactors, test-only changes, CI config, dependency bumps, doc-only
+changes, style/formatting commits. These don't affect doc accuracy.
+
+For borderline cases, check the actual diff — a commit titled "refactor: X"
+that adds a new public API is a feature.
+
+### Step 3 — Build a change summary
+
+Produce a concise list like:
+
+```
+Since last review (<sha>, <date>):
+- FEATURE: Plugin system merged (runtime, SDK, CLI, slots, event bridge)
+- FEATURE: Project archiving added
+- BREAKING: Removed legacy webhook adapter
+- STRUCTURAL: New .agents/skills/ directory convention
+```
+
+If there are no notable changes, skip to Step 7 (update cursor and exit).
+
+### Step 4 — Audit each target doc
+
+For each target document, read it fully and cross-reference against the change
+summary. Check for:
+
+1. **False negatives** — major shipped features not mentioned at all
+2. **False positives** — features listed as "coming soon" / "roadmap" / "planned"
+   / "not supported" / "TBD" that already shipped
+3. **Quickstart accuracy** — install commands, prereqs, and startup instructions
+   still correct (README only)
+4. **Feature table accuracy** — does the features section reflect current
+   capabilities? (README only)
+5. **Works-with accuracy** — are supported adapters/integrations listed correctly?
+
+Use `references/audit-checklist.md` as the structured checklist.
+Use `references/section-map.md` to know where to look for each feature area.
+
+### Step 5 — Create branch and apply minimal edits
+
+```bash
+# Create a branch for the doc updates
+BRANCH="docs/maintenance-$(date +%Y%m%d)"
+git checkout -b "$BRANCH"
+```
+
+Apply **only** the edits needed to fix drift. Rules:
+
+- **Minimal patches only.** Fix inaccuracies, don't rewrite sections.
+- **Preserve voice and style.** Match the existing tone of each document.
+- **No cosmetic changes.** Don't fix typos, reformat tables, or reorganize
+  sections unless they're part of a factual fix.
+- **No new sections.** If a feature needs a whole new section, note it in the
+  PR description as a follow-up — don't add it in a maintenance pass.
+- **Roadmap items:** Move shipped features out of Roadmap. Add a brief mention
+  in the appropriate existing section if there isn't one already. Don't add
+  long descriptions.
+
+### Step 6 — Open a PR
+
+Commit the changes and open a PR:
+
+```bash
+git add README.md doc/SPEC.md doc/PRODUCT.md .doc-review-cursor
+git commit -m "docs: update documentation for accuracy
+
+- [list each fix briefly]
+
+Co-Authored-By: Paperclip <noreply@paperclip.ing>"
+
+git push -u origin "$BRANCH"
+
+gh pr create \
+  --title "docs: periodic documentation accuracy update" \
+  --body "$(cat <<'EOF'
+## Summary
+Automated doc maintenance pass. Fixes documentation drift detected since
+last review.
+
+### Changes
+- [list each fix]
+
+### Change summary (since last review)
+- [list notable code changes that triggered doc updates]
+
+## Review notes
+- Only factual accuracy fixes — no style/cosmetic changes
+- Preserves existing voice and structure
+- Larger doc additions (new sections, tutorials) noted as follow-ups
+
+🤖 Generated by doc-maintenance skill
+EOF
+)"
+```
+
+### Step 7 — Update the cursor
+
+After a successful audit (whether or not edits were needed), update the cursor:
+
+```bash
+git rev-parse HEAD > .doc-review-cursor
+```
+
+If edits were made, this is already committed in the PR branch. If no edits
+were needed, commit the cursor update to the current branch.
+
+## Change Classification Rules
+
+| Signal | Category | Doc update needed? |
+|--------|----------|-------------------|
+| `feat:`, `add`, `implement`, `support` in message | Feature | Yes if user-facing |
+| `remove`, `drop`, `breaking`, `!:` in message | Breaking | Yes |
+| New top-level directory or config file | Structural | Maybe |
+| `fix:`, `bugfix` | Fix | No (unless it changes behavior described in docs) |
+| `refactor:`, `chore:`, `ci:`, `test:` | Maintenance | No |
+| `docs:` | Doc change | No (already handled) |
+| Dependency bumps only | Maintenance | No |
+
+## Patch Style Guide
+
+- Fix the fact, not the prose
+- If removing a roadmap item, don't leave a gap — remove the bullet cleanly
+- If adding a feature mention, match the format of surrounding entries
+  (e.g. if features are in a table, add a table row)
+- Keep README changes especially minimal — it shouldn't churn often
+- For SPEC/PRODUCT, prefer updating existing statements over adding new ones
+  (e.g. change "not supported in V1" to "supported via X" rather than adding
+  a new section)
+
+## Output
+
+When the skill completes, report:
+
+- How many commits were scanned
+- How many notable changes were found
+- How many doc edits were made (and to which files)
+- PR link (if edits were made)
+- Any follow-up items that need larger doc work

.agents/skills/doc-maintenance/references/audit-checklist.md

@@ -0,0 +1,85 @@
+# Doc Maintenance Audit Checklist
+
+Use this checklist when auditing each target document. For each item, compare
+against the change summary from git history.
+
+## README.md
+
+### Features table
+- [ ] Each feature card reflects a shipped capability
+- [ ] No feature cards for things that don't exist yet
+- [ ] No major shipped features missing from the table
+
+### Roadmap
+- [ ] Nothing listed as "planned" or "coming soon" that already shipped
+- [ ] No removed/cancelled items still listed
+- [ ] Items reflect current priorities (cross-check with recent PRs)
+
+### Quickstart
+- [ ] `npx paperclipai onboard` command is correct
+- [ ] Manual install steps are accurate (clone URL, commands)
+- [ ] Prerequisites (Node version, pnpm version) are current
+- [ ] Server URL and port are correct
+
+### "What is Paperclip" section
+- [ ] High-level description is accurate
+- [ ] Step table (Define goal / Hire team / Approve and run) is correct
+
+### "Works with" table
+- [ ] All supported adapters/runtimes are listed
+- [ ] No removed adapters still listed
+- [ ] Logos and labels match current adapter names
+
+### "Paperclip is right for you if"
+- [ ] Use cases are still accurate
+- [ ] No claims about capabilities that don't exist
+
+### "Why Paperclip is special"
+- [ ] Technical claims are accurate (atomic execution, governance, etc.)
+- [ ] No features listed that were removed or significantly changed
+
+### FAQ
+- [ ] Answers are still correct
+- [ ] No references to removed features or outdated behavior
+
+### Development section
+- [ ] Commands are accurate (`pnpm dev`, `pnpm build`, etc.)
+- [ ] Link to DEVELOPING.md is correct
+
+## doc/SPEC.md
+
+### Company Model
+- [ ] Fields match current schema
+- [ ] Governance model description is accurate
+
+### Agent Model
+- [ ] Adapter types match what's actually supported
+- [ ] Agent configuration description is accurate
+- [ ] No features described as "not supported" or "not V1" that shipped
+
+### Task Model
+- [ ] Task hierarchy description is accurate
+- [ ] Status values match current implementation
+
+### Extensions / Plugins
+- [ ] If plugins are shipped, no "not in V1" or "future" language
+- [ ] Plugin model description matches implementation
+
+### Open Questions
+- [ ] Resolved questions removed or updated
+- [ ] No "TBD" items that have been decided
+
+## doc/PRODUCT.md
+
+### Core Concepts
+- [ ] Company, Employees, Task Management descriptions accurate
+- [ ] Agent Execution modes described correctly
+- [ ] No missing major concepts
+
+### Principles
+- [ ] Principles haven't been contradicted by shipped features
+- [ ] No principles referencing removed capabilities
+
+### User Flow
+- [ ] Dream scenario still reflects actual onboarding
+- [ ] Steps are achievable with current features

.agents/skills/doc-maintenance/references/section-map.md

@@ -0,0 +1,22 @@
+# Section Map
+
+Maps feature areas to specific document sections so the skill knows where to
+look when a feature ships or changes.
+
+| Feature Area | README Section | SPEC Section | PRODUCT Section |
+|-------------|---------------|-------------|----------------|
+| Plugins / Extensions | Features table, Roadmap | Extensions, Agent Model | Core Concepts |
+| Adapters (new runtimes) | "Works with" table, FAQ | Agent Model, Agent Configuration | Employees & Agents, Agent Execution |
+| Governance / Approvals | Features table, "Why special" | Board Governance, Board Approval Gates | Principles |
+| Budget / Cost Control | Features table, "Why special" | Budget Delegation | Company (revenue & expenses) |
+| Task Management | Features table | Task Model | Task Management |
+| Org Chart / Hierarchy | Features table | Agent Model (reporting) | Employees & Agents |
+| Multi-Company | Features table, FAQ | Company Model | Company |
+| Heartbeats | Features table, FAQ | Agent Execution | Agent Execution |
+| CLI Commands | Development section | — | — |
+| Onboarding / Quickstart | Quickstart, FAQ | — | User Flow |
+| Skills / Skill Injection | "Why special" | — | — |
+| Company Templates | "Why special", Roadmap (ClipMart) | — | — |
+| Mobile / UI | Features table | — | — |
+| Project Archiving | — | — | — |
+| OpenClaw Integration | "Works with" table, FAQ | Agent Model | Agent Execution |

.agents/skills/pr-report/SKILL.md

@@ -0,0 +1,202 @@
+---
+name: pr-report
+description: >
+  Review a pull request or contribution deeply, explain it tutorial-style for a
+  maintainer, and produce a polished report artifact such as HTML or Markdown.
+  Use when asked to analyze a PR, explain a contributor's design decisions,
+  compare it with similar systems, or prepare a merge recommendation.
+---
+
+# PR Report Skill
+
+Produce a maintainer-grade review of a PR, branch, or large contribution.
+
+Default posture:
+
+- understand the change before judging it
+- explain the system as built, not just the diff
+- separate architectural problems from product-scope objections
+- make a concrete recommendation, not a vague impression
+
+## When to Use
+
+Use this skill when the user asks for things like:
+
+- "review this PR deeply"
+- "explain this contribution to me"
+- "make me a report or webpage for this PR"
+- "compare this design to similar systems"
+- "should I merge this?"
+
+## Outputs
+
+Common outputs:
+
+- standalone HTML report in `tmp/reports/...`
+- Markdown report in `report/` or another requested folder
+- short maintainer summary in chat
+
+If the user asks for a webpage, build a polished standalone HTML artifact with
+clear sections and readable visual hierarchy.
+
+Resources bundled with this skill:
+
+- `references/style-guide.md` for visual direction and report presentation rules
+- `assets/html-report-starter.html` for a reusable standalone HTML/CSS starter
+
+## Workflow
+
+### 1. Acquire and frame the target
+
+Work from local code when possible, not just the GitHub PR page.
+
+Gather:
+
+- target branch or worktree
+- diff size and changed subsystems
+- relevant repo docs, specs, and invariants
+- contributor intent if it is documented in PR text or design docs
+
+Start by answering: what is this change *trying* to become?
+
+### 2. Build a mental model of the system
+
+Do not stop at file-by-file notes. Reconstruct the design:
+
+- what new runtime or contract exists
+- which layers changed: db, shared types, server, UI, CLI, docs
+- lifecycle: install, startup, execution, UI, failure, disablement
+- trust boundary: what code runs where, under what authority
+
+For large contributions, include a tutorial-style section that teaches the
+system from first principles.
+
+### 3. Review like a maintainer
+
+Findings come first. Order by severity.
+
+Prioritize:
+
+- behavioral regressions
+- trust or security gaps
+- misleading abstractions
+- lifecycle and operational risks
+- coupling that will be hard to unwind
+- missing tests or unverifiable claims
+
+Always cite concrete file references when possible.
+
+### 4. Distinguish the objection type
+
+Be explicit about whether a concern is:
+
+- product direction
+- architecture
+- implementation quality
+- rollout strategy
+- documentation honesty
+
+Do not hide an architectural objection inside a scope objection.
+
+### 5. Compare to external precedents when needed
+
+If the contribution introduces a framework or platform concept, compare it to
+similar open-source systems.
+
+When comparing:
+
+- prefer official docs or source
+- focus on extension boundaries, context passing, trust model, and UI ownership
+- extract lessons, not just similarities
+
+Good comparison questions:
+
+- Who owns lifecycle?
+- Who owns UI composition?
+- Is context explicit or ambient?
+- Are plugins trusted code or sandboxed code?
+- Are extension points named and typed?
+
+### 6. Make the recommendation actionable
+
+Do not stop at "merge" or "do not merge."
+
+Choose one:
+
+- merge as-is
+- merge after specific redesign
+- salvage specific pieces
+- keep as design research
+
+If rejecting or narrowing, say what should be kept.
+
+Useful recommendation buckets:
+
+- keep the protocol/type model
+- redesign the UI boundary
+- narrow the initial surface area
+- defer third-party execution
+- ship a host-owned extension-point model first
+
+### 7. Build the artifact
+
+Suggested report structure:
+
+1. Executive summary
+2. What the PR actually adds
+3. Tutorial: how the system works
+4. Strengths
+5. Main findings
+6. Comparisons
+7. Recommendation
+
+For HTML reports:
+
+- use intentional typography and color
+- make navigation easy for long reports
+- favor strong section headings and small reference labels
+- avoid generic dashboard styling
+
+Before building from scratch, read `references/style-guide.md`.
+If a fast polished starter is helpful, begin from `assets/html-report-starter.html`
+and replace the placeholder content with the actual report.
+
+### 8. Verify before handoff
+
+Check:
+
+- artifact path exists
+- findings still match the actual code
+- any requested forbidden strings are absent from generated output
+- if tests were not run, say so explicitly
+
+## Review Heuristics
+
+### Plugin and platform work
+
+Watch closely for:
+
+- docs claiming sandboxing while runtime executes trusted host processes
+- module-global state used to smuggle React context
+- hidden dependence on render order
+- plugins reaching into host internals instead of using explicit APIs
+- "capabilities" that are really policy labels on top of fully trusted code
+
+### Good signs
+
+- typed contracts shared across layers
+- explicit extension points
+- host-owned lifecycle
+- honest trust model
+- narrow first rollout with room to grow
+
+## Final Response
+
+In chat, summarize:
+
+- where the report is
+- your overall call
+- the top one or two reasons
+- whether verification or tests were skipped
+
+Keep the chat summary shorter than the report itself.

.agents/skills/pr-report/assets/html-report-starter.html

@@ -0,0 +1,426 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1" />
+    <title>PR Report Starter</title>
+    <link rel="preconnect" href="https://fonts.googleapis.com" />
+    <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
+    <link
+      href="https://fonts.googleapis.com/css2?family=IBM+Plex+Sans:wght@400;500;600;700&family=Newsreader:opsz,wght@6..72,500;6..72,700&display=swap"
+      rel="stylesheet"
+    />
+    <style>
+      :root {
+        --bg: #f4efe5;
+        --paper: rgba(255, 251, 244, 0.88);
+        --paper-strong: #fffaf1;
+        --ink: #1f1b17;
+        --muted: #6a6257;
+        --line: rgba(31, 27, 23, 0.12);
+        --accent: #9c4729;
+        --accent-soft: rgba(156, 71, 41, 0.1);
+        --good: #2f6a42;
+        --warn: #946200;
+        --bad: #8c2f25;
+        --shadow: 0 22px 60px rgba(52, 37, 19, 0.1);
+        --radius: 20px;
+      }
+
+      * {
+        box-sizing: border-box;
+      }
+
+      html {
+        scroll-behavior: smooth;
+      }
+
+      body {
+        margin: 0;
+        color: var(--ink);
+        font-family: "IBM Plex Sans", sans-serif;
+        background:
+          radial-gradient(circle at top left, rgba(156, 71, 41, 0.12), transparent 34rem),
+          radial-gradient(circle at top right, rgba(47, 106, 66, 0.08), transparent 28rem),
+          linear-gradient(180deg, #efe6d6 0%, var(--bg) 48%, #ece5d8 100%);
+      }
+
+      .shell {
+        width: min(1360px, calc(100vw - 32px));
+        margin: 24px auto;
+        display: grid;
+        grid-template-columns: 280px minmax(0, 1fr);
+        gap: 24px;
+      }
+
+      .panel {
+        background: var(--paper);
+        backdrop-filter: blur(12px);
+        border: 1px solid var(--line);
+        border-radius: var(--radius);
+        box-shadow: var(--shadow);
+      }
+
+      .nav {
+        position: sticky;
+        top: 20px;
+        align-self: start;
+        padding: 22px;
+      }
+
+      .eyebrow {
+        letter-spacing: 0.12em;
+        text-transform: uppercase;
+        font-size: 11px;
+        font-weight: 700;
+        color: var(--accent);
+      }
+
+      .nav h1,
+      .hero h1,
+      h2,
+      h3 {
+        font-family: "Newsreader", serif;
+        line-height: 0.96;
+        margin: 0;
+      }
+
+      .nav h1 {
+        font-size: 2rem;
+        margin-top: 10px;
+      }
+
+      .nav p {
+        color: var(--muted);
+        font-size: 0.95rem;
+        line-height: 1.5;
+      }
+
+      .nav ul {
+        list-style: none;
+        padding: 0;
+        margin: 18px 0 0;
+        display: grid;
+        gap: 10px;
+      }
+
+      .nav a {
+        display: block;
+        color: var(--ink);
+        text-decoration: none;
+        padding: 10px 12px;
+        border-radius: 12px;
+        border: 1px solid transparent;
+        background: rgba(255, 255, 255, 0.35);
+      }
+
+      .nav a:hover {
+        border-color: var(--line);
+        background: rgba(255, 255, 255, 0.75);
+      }
+
+      .meta-block {
+        margin-top: 20px;
+        padding-top: 18px;
+        border-top: 1px solid var(--line);
+        color: var(--muted);
+        font-size: 0.86rem;
+        line-height: 1.5;
+      }
+
+      main {
+        display: grid;
+        gap: 24px;
+      }
+
+      section {
+        padding: 26px 28px 28px;
+      }
+
+      .hero {
+        padding: 28px;
+        overflow: hidden;
+        position: relative;
+      }
+
+      .hero::after {
+        content: "";
+        position: absolute;
+        inset: auto -3rem -6rem auto;
+        width: 18rem;
+        height: 18rem;
+        border-radius: 50%;
+        background: radial-gradient(circle, rgba(156, 71, 41, 0.14), transparent 68%);
+        pointer-events: none;
+      }
+
+      .hero h1 {
+        font-size: clamp(2.6rem, 5vw, 4.6rem);
+        max-width: 12ch;
+        margin-top: 12px;
+      }
+
+      .lede {
+        margin-top: 16px;
+        max-width: 70ch;
+        font-size: 1.05rem;
+        line-height: 1.65;
+        color: #2b2723;
+      }
+
+      .hero-grid,
+      .card-grid,
+      .two-col {
+        display: grid;
+        gap: 14px;
+      }
+
+      .hero-grid {
+        margin-top: 24px;
+        grid-template-columns: repeat(4, minmax(0, 1fr));
+      }
+
+      .card-grid {
+        grid-template-columns: repeat(2, minmax(0, 1fr));
+      }
+
+      .two-col {
+        grid-template-columns: repeat(2, minmax(0, 1fr));
+      }
+
+      .metric,
+      .card,
+      .finding {
+        padding: 18px;
+        background: rgba(255, 255, 255, 0.68);
+        border: 1px solid var(--line);
+        border-radius: 18px;
+      }
+
+      .metric .label {
+        color: var(--muted);
+        font-size: 0.82rem;
+        text-transform: uppercase;
+        letter-spacing: 0.08em;
+      }
+
+      .metric .value {
+        margin-top: 8px;
+        font-size: 1.45rem;
+        font-weight: 700;
+      }
+
+      h2 {
+        font-size: 2rem;
+        margin-bottom: 16px;
+      }
+
+      h3 {
+        font-size: 1.3rem;
+        margin-bottom: 10px;
+      }
+
+      p {
+        margin: 0 0 14px;
+        line-height: 1.65;
+      }
+
+      ul,
+      ol {
+        margin: 0;
+        padding-left: 20px;
+        line-height: 1.65;
+      }
+
+      li + li {
+        margin-top: 8px;
+      }
+
+      .badge-row {
+        display: flex;
+        flex-wrap: wrap;
+        gap: 8px;
+        margin: 18px 0 8px;
+      }
+
+      .badge {
+        display: inline-flex;
+        align-items: center;
+        gap: 8px;
+        padding: 8px 10px;
+        border-radius: 999px;
+        font-size: 0.82rem;
+        font-weight: 700;
+        border: 1px solid var(--line);
+        background: rgba(255, 255, 255, 0.68);
+      }
+
+      .badge.good {
+        color: var(--good);
+      }
+
+      .badge.warn {
+        color: var(--warn);
+      }
+
+      .badge.bad {
+        color: var(--bad);
+      }
+
+      .quote {
+        margin-top: 18px;
+        padding: 18px;
+        border-left: 4px solid var(--accent);
+        border-radius: 14px;
+        background: var(--accent-soft);
+      }
+
+      .severity {
+        display: inline-flex;
+        margin-bottom: 12px;
+        padding: 6px 10px;
+        border-radius: 999px;
+        font-size: 0.78rem;
+        font-weight: 700;
+        text-transform: uppercase;
+        letter-spacing: 0.08em;
+      }
+
+      .severity.high {
+        background: rgba(140, 47, 37, 0.12);
+        color: var(--bad);
+      }
+
+      .severity.medium {
+        background: rgba(148, 98, 0, 0.12);
+        color: var(--warn);
+      }
+
+      .severity.low {
+        background: rgba(47, 106, 66, 0.12);
+        color: var(--good);
+      }
+
+      .ref {
+        color: var(--muted);
+        font-size: 0.82rem;
+        line-height: 1.5;
+      }
+
+      @media (max-width: 980px) {
+        .shell {
+          grid-template-columns: 1fr;
+        }
+
+        .nav {
+          position: static;
+        }
+
+        .hero-grid,
+        .card-grid,
+        .two-col {
+          grid-template-columns: 1fr;
+        }
+
+        .hero h1 {
+          max-width: 100%;
+        }
+      }
+    </style>
+  </head>
+  <body>
+    <div class="shell">
+      <aside class="panel nav">
+        <div class="eyebrow">Maintainer Report</div>
+        <h1>Report Title</h1>
+        <p>Replace this with a concise description of what the report covers.</p>
+        <ul>
+          <li><a href="#summary">Summary</a></li>
+          <li><a href="#tutorial">Tutorial</a></li>
+          <li><a href="#findings">Findings</a></li>
+          <li><a href="#recommendation">Recommendation</a></li>
+        </ul>
+        <div class="meta-block">
+          Replace with project metadata, review date, or scope notes.
+        </div>
+      </aside>
+
+      <main>
+        <section class="panel hero" id="summary">
+          <div class="eyebrow">Executive Summary</div>
+          <h1>Use the hero for the clearest one-line judgment.</h1>
+          <p class="lede">
+            Replace this with the short explanation of what the contribution does, why it matters,
+            and what the core maintainer question is.
+          </p>
+          <div class="badge-row">
+            <span class="badge good">Strength</span>
+            <span class="badge warn">Tradeoff</span>
+            <span class="badge bad">Risk</span>
+          </div>
+          <div class="hero-grid">
+            <div class="metric">
+              <div class="label">Overall Call</div>
+              <div class="value">Placeholder</div>
+            </div>
+            <div class="metric">
+              <div class="label">Main Concern</div>
+              <div class="value">Placeholder</div>
+            </div>
+            <div class="metric">
+              <div class="label">Best Part</div>
+              <div class="value">Placeholder</div>
+            </div>
+            <div class="metric">
+              <div class="label">Weakest Part</div>
+              <div class="value">Placeholder</div>
+            </div>
+          </div>
+          <div class="quote">
+            Use this block for the thesis, a sharp takeaway, or a key cited point.
+          </div>
+        </section>
+
+        <section class="panel" id="tutorial">
+          <h2>Tutorial Section</h2>
+          <div class="two-col">
+            <div class="card">
+              <h3>Concept Card</h3>
+              <p>Use cards for mental models, subsystems, or comparison slices.</p>
+              <div class="ref">path/to/file.ts:10</div>
+            </div>
+            <div class="card">
+              <h3>Second Card</h3>
+              <p>Keep cards fairly dense. This template is about style, not fixed structure.</p>
+              <div class="ref">path/to/file.ts:20</div>
+            </div>
+          </div>
+        </section>
+
+        <section class="panel" id="findings">
+          <h2>Findings</h2>
+          <article class="finding">
+            <div class="severity high">High</div>
+            <h3>Finding Title</h3>
+            <p>Use findings for the sharpest judgment calls and risks.</p>
+            <div class="ref">path/to/file.ts:30</div>
+          </article>
+        </section>
+
+        <section class="panel" id="recommendation">
+          <h2>Recommendation</h2>
+          <div class="card-grid">
+            <div class="card">
+              <h3>Path Forward</h3>
+              <p>Use this area for merge guidance, salvage plan, or rollout advice.</p>
+            </div>
+            <div class="card">
+              <h3>What To Keep</h3>
+              <p>Call out the parts worth preserving even if the whole proposal should not land.</p>
+            </div>
+          </div>
+        </section>
+      </main>
+    </div>
+  </body>
+</html>

.agents/skills/pr-report/references/style-guide.md

@@ -0,0 +1,149 @@
+# PR Report Style Guide
+
+Use this guide when the user wants a report artifact, especially a webpage.
+
+## Goal
+
+Make the report feel like an editorial review, not an internal admin dashboard.
+The page should make a long technical argument easy to scan without looking
+generic or overdesigned.
+
+## Visual Direction
+
+Preferred tone:
+
+- editorial
+- warm
+- serious
+- high-contrast
+- handcrafted, not corporate SaaS
+
+Avoid:
+
+- default app-shell layouts
+- purple gradients on white
+- generic card dashboards
+- cramped pages with weak hierarchy
+- novelty fonts that hurt readability
+
+## Typography
+
+Recommended pattern:
+
+- one expressive serif or display face for major headings
+- one sturdy sans-serif for body copy and UI labels
+
+Good combinations:
+
+- Newsreader + IBM Plex Sans
+- Source Serif 4 + Instrument Sans
+- Fraunces + Public Sans
+- Libre Baskerville + Work Sans
+
+Rules:
+
+- headings should feel deliberate and large
+- body copy should stay comfortable for long reading
+- reference labels and badges should use smaller dense sans text
+
+## Layout
+
+Recommended structure:
+
+- a sticky side or top navigation for long reports
+- one strong hero summary at the top
+- panel or paper-like sections for each major topic
+- multi-column card grids for comparisons and strengths
+- single-column body text for findings and recommendations
+
+Use generous spacing. Long-form technical reports need breathing room.
+
+## Color
+
+Prefer muted paper-like backgrounds with one warm accent and one cool counterweight.
+
+Suggested token categories:
+
+- `--bg`
+- `--paper`
+- `--ink`
+- `--muted`
+- `--line`
+- `--accent`
+- `--good`
+- `--warn`
+- `--bad`
+
+The accent should highlight navigation, badges, and important labels. Do not
+let accent colors dominate body text.
+
+## Useful UI Elements
+
+Include small reusable styles for:
+
+- summary metrics
+- badges
+- quotes or callouts
+- finding cards
+- severity labels
+- reference labels
+- comparison cards
+- responsive two-column sections
+
+## Motion
+
+Keep motion restrained.
+
+Good:
+
+- soft fade/slide-in on first load
+- hover response on nav items or cards
+
+Bad:
+
+- constant animation
+- floating blobs
+- decorative motion with no reading benefit
+
+## Content Presentation
+
+Even when the user wants design polish, clarity stays primary.
+
+Good structure for long reports:
+
+1. executive summary
+2. what changed
+3. tutorial explanation
+4. strengths
+5. findings
+6. comparisons
+7. recommendation
+
+The exact headings can change. The important thing is to separate explanation
+from judgment.
+
+## References
+
+Reference labels should be visually quiet but easy to spot.
+
+Good pattern:
+
+- small muted text
+- monospace or compact sans
+- keep them close to the paragraph they support
+
+## Starter Usage
+
+If you need a fast polished base, start from:
+
+- `assets/html-report-starter.html`
+
+Customize:
+
+- fonts
+- color tokens
+- hero copy
+- section ordering
+- card density
+
+Do not preserve the placeholder sections if they do not fit the actual report.

.agents/skills/release-changelog/SKILL.md

@@ -0,0 +1,192 @@
+---
+name: release-changelog
+description: >
+  Generate the stable Paperclip release changelog at releases/vYYYY.MDD.P.md by
+  reading commits, changesets, and merged PR context since the last stable tag.
+---
+
+# Release Changelog Skill
+
+Generate the user-facing changelog for the **stable** Paperclip release.
+
+## Versioning Model
+
+Paperclip uses **calendar versioning (calver)**:
+
+- Stable releases: `YYYY.MDD.P` (e.g. `2026.318.0`)
+- Canary releases: `YYYY.MDD.P-canary.N` (e.g. `2026.318.1-canary.0`)
+- Git tags: `vYYYY.MDD.P` for stable, `canary/vYYYY.MDD.P-canary.N` for canary
+
+There are no major/minor/patch bumps. The stable version is derived from the
+intended release date (UTC) plus the next same-day stable patch slot.
+
+Output:
+
+- `releases/vYYYY.MDD.P.md`
+
+Important rules:
+
+- even if there are canary releases such as `2026.318.1-canary.0`, the changelog file stays `releases/v2026.318.1.md`
+- do not derive versions from semver bump types
+- do not create canary changelog files
+
+## Step 0 — Idempotency Check
+
+Before generating anything, check whether the file already exists:
+
+```bash
+ls releases/vYYYY.MDD.P.md 2>/dev/null
+```
+
+If it exists:
+
+1. read it first
+2. present it to the reviewer
+3. ask whether to keep it, regenerate it, or update specific sections
+4. never overwrite it silently
+
+## Step 1 — Determine the Stable Range
+
+Find the last stable tag:
+
+```bash
+git tag --list 'v*' --sort=-version:refname | head -1
+git log v{last}..HEAD --oneline --no-merges
+```
+
+The stable version comes from one of:
+
+- an explicit maintainer request
+- `./scripts/release.sh stable --date YYYY-MM-DD --print-version`
+- the release plan already agreed in `doc/RELEASING.md`
+
+Do not derive the changelog version from a canary tag or prerelease suffix.
+Do not derive major/minor/patch bumps from API intent — calver uses the date and same-day stable slot.
+
+## Step 2 — Gather the Raw Inputs
+
+Collect release data from:
+
+1. git commits since the last stable tag
+2. `.changeset/*.md` files
+3. merged PRs via `gh` when available
+
+Useful commands:
+
+```bash
+git log v{last}..HEAD --oneline --no-merges
+git log v{last}..HEAD --format="%H %s" --no-merges
+ls .changeset/*.md | grep -v README.md
+gh pr list --state merged --search "merged:>={last-tag-date}" --json number,title,body,labels
+```
+
+## Step 3 — Detect Breaking Changes
+
+Look for:
+
+- destructive migrations
+- removed or changed API fields/endpoints
+- renamed or removed config keys
+- `BREAKING:` or `BREAKING CHANGE:` commit signals
+
+Key commands:
+
+```bash
+git diff --name-only v{last}..HEAD -- packages/db/src/migrations/
+git diff v{last}..HEAD -- packages/db/src/schema/
+git diff v{last}..HEAD -- server/src/routes/ server/src/api/
+git log v{last}..HEAD --format="%s" | rg -n 'BREAKING CHANGE|BREAKING:|^[a-z]+!:' || true
+```
+
+If breaking changes are detected, flag them prominently — they must appear in the
+Breaking Changes section with an upgrade path.
+
+## Step 4 — Categorize for Users
+
+Use these stable changelog sections:
+
+- `Breaking Changes`
+- `Highlights`
+- `Improvements`
+- `Fixes`
+- `Upgrade Guide` when needed
+
+Exclude purely internal refactors, CI changes, and docs-only work unless they materially affect users.
+
+Guidelines:
+
+- group related commits into one user-facing entry
+- write from the user perspective
+- keep highlights short and concrete
+- spell out upgrade actions for breaking changes
+
+### Inline PR and contributor attribution
+
+When a bullet item clearly maps to a merged pull request, add inline attribution at the
+end of the entry in this format:
+
+```
+- **Feature name** — Description. ([#123](https://github.com/paperclipai/paperclip/pull/123), @contributor1, @contributor2)
+```
+
+Rules:
+
+- Only add a PR link when you can confidently trace the bullet to a specific merged PR.
+  Use merge commit messages (`Merge pull request #N from user/branch`) to map PRs.
+- List the contributor(s) who authored the PR. Use GitHub usernames, not real names or emails.
+- If multiple PRs contributed to a single bullet, list them all: `([#10](url), [#12](url), @user1, @user2)`.
+- If you cannot determine the PR number or contributor with confidence, omit the attribution
+  parenthetical — do not guess.
+- Core maintainer commits that don't have an external PR can omit the parenthetical.
+
+## Step 5 — Write the File
+
+Template:
+
+```markdown
+# vYYYY.MDD.P
+
+> Released: YYYY-MM-DD
+
+## Breaking Changes
+
+## Highlights
+
+## Improvements
+
+## Fixes
+
+## Upgrade Guide
+
+## Contributors
+
+Thank you to everyone who contributed to this release!
+
+@username1, @username2, @username3
+```
+
+Omit empty sections except `Highlights`, `Improvements`, and `Fixes`, which should usually exist.
+
+The `Contributors` section should always be included. List every person who authored
+commits in the release range, @-mentioning them by their **GitHub username** (not their
+real name or email). To find GitHub usernames:
+
+1. Extract usernames from merge commit messages: `git log v{last}..HEAD --oneline --merges` — the branch prefix (e.g. `from username/branch`) gives the GitHub username.
+2. For noreply emails like `user@users.noreply.github.com`, the username is the part before `@`.
+3. For contributors whose username is ambiguous, check `gh api users/{guess}` or the PR page.
+
+**Never expose contributor email addresses.** Use `@username` only.
+
+Exclude bot accounts (e.g. `lockfile-bot`, `dependabot`) from the list. List contributors
+in alphabetical order by GitHub username (case-insensitive).
+
+## Step 6 — Review Before Release
+
+Before handing it off:
+
+1. confirm the heading is the stable version only
+2. confirm there is no `-canary` language in the title or filename
+3. confirm any breaking changes have an upgrade path
+4. present the draft for human sign-off
+
+This skill never publishes anything. It only prepares the stable changelog artifact.

.agents/skills/release/SKILL.md

@@ -0,0 +1,247 @@
+---
+name: release
+description: >
+  Coordinate a full Paperclip release across engineering verification, npm,
+  GitHub, smoke testing, and announcement follow-up. Use when leadership asks
+  to ship a release, not merely to discuss versioning.
+---
+
+# Release Coordination Skill
+
+Run the full Paperclip maintainer release workflow, not just an npm publish.
+
+This skill coordinates:
+
+- stable changelog drafting via `release-changelog`
+- canary verification and publish status from `master`
+- Docker smoke testing via `scripts/docker-onboard-smoke.sh`
+- manual stable promotion from a chosen source ref
+- GitHub Release creation
+- website / announcement follow-up tasks
+
+## Trigger
+
+Use this skill when leadership asks for:
+
+- "do a release"
+- "ship the release"
+- "promote this canary to stable"
+- "cut the stable release"
+
+## Preconditions
+
+Before proceeding, verify all of the following:
+
+1. `.agents/skills/release-changelog/SKILL.md` exists and is usable.
+2. The repo working tree is clean, including untracked files.
+3. There is at least one canary or candidate commit since the last stable tag.
+4. The candidate SHA has passed the verification gate or is about to.
+5. If manifests changed, the CI-owned `pnpm-lock.yaml` refresh is already merged on `master`.
+6. npm publish rights are available through GitHub trusted publishing, or through local npm auth for emergency/manual use.
+7. If running through Paperclip, you have issue context for status updates and follow-up task creation.
+
+If any precondition fails, stop and report the blocker.
+
+## Inputs
+
+Collect these inputs up front:
+
+- whether the target is a canary check or a stable promotion
+- the candidate `source_ref` for stable
+- whether the stable run is dry-run or live
+- release issue / company context for website and announcement follow-up
+
+## Step 0 — Release Model
+
+Paperclip now uses a commit-driven release model:
+
+1. every push to `master` publishes a canary automatically
+2. canaries use `YYYY.MDD.P-canary.N`
+3. stable releases use `YYYY.MDD.P`
+4. the middle slot is `MDD`, where `M` is the UTC month and `DD` is the zero-padded UTC day
+5. the stable patch slot increments when more than one stable ships on the same UTC date
+6. stable releases are manually promoted from a chosen tested commit or canary source commit
+7. only stable releases get `releases/vYYYY.MDD.P.md`, git tag `vYYYY.MDD.P`, and a GitHub Release
+
+Critical consequences:
+
+- do not use release branches as the default path
+- do not derive major/minor/patch bumps
+- do not create canary changelog files
+- do not create canary GitHub Releases
+
+## Step 1 — Choose the Candidate
+
+For canary validation:
+
+- inspect the latest successful canary run on `master`
+- record the canary version and source SHA
+
+For stable promotion:
+
+1. choose the tested source ref
+2. confirm it is the exact SHA you want to promote
+3. resolve the target stable version with `./scripts/release.sh stable --date YYYY-MM-DD --print-version`
+
+Useful commands:
+
+```bash
+git tag --list 'v*' --sort=-version:refname | head -1
+git log --oneline --no-merges
+npm view paperclipai@canary version
+```
+
+## Step 2 — Draft the Stable Changelog
+
+Stable changelog files live at:
+
+- `releases/vYYYY.MDD.P.md`
+
+Invoke `release-changelog` and generate or update the stable notes only.
+
+Rules:
+
+- review the draft with a human before publish
+- preserve manual edits if the file already exists
+- keep the filename stable-only
+- do not create a canary changelog file
+
+## Step 3 — Verify the Candidate SHA
+
+Run the standard gate:
+
+```bash
+pnpm -r typecheck
+pnpm test:run
+pnpm build
+```
+
+If the GitHub release workflow will run the publish, it can rerun this gate. Still report local status if you checked it.
+
+For PRs that touch release logic, the repo also runs a canary release dry-run in CI. That is a release-specific guard, not a substitute for the standard gate.
+
+## Step 4 — Validate the Canary
+
+The normal canary path is automatic from `master` via:
+
+- `.github/workflows/release.yml`
+
+Confirm:
+
+1. verification passed
+2. npm canary publish succeeded
+3. git tag `canary/vYYYY.MDD.P-canary.N` exists
+
+Useful checks:
+
+```bash
+npm view paperclipai@canary version
+git tag --list 'canary/v*' --sort=-version:refname | head -5
+```
+
+## Step 5 — Smoke Test the Canary
+
+Run:
+
+```bash
+PAPERCLIPAI_VERSION=canary ./scripts/docker-onboard-smoke.sh
+```
+
+Useful isolated variant:
+
+```bash
+HOST_PORT=3232 DATA_DIR=./data/release-smoke-canary PAPERCLIPAI_VERSION=canary ./scripts/docker-onboard-smoke.sh
+```
+
+Confirm:
+
+1. install succeeds
+2. onboarding completes without crashes
+3. the server boots
+4. the UI loads
+5. basic company creation and dashboard load work
+
+If smoke testing fails:
+
+- stop the stable release
+- fix the issue on `master`
+- wait for the next automatic canary
+- rerun smoke testing
+
+## Step 6 — Preview or Publish Stable
+
+The normal stable path is manual `workflow_dispatch` on:
+
+- `.github/workflows/release.yml`
+
+Inputs:
+
+- `source_ref`
+- `stable_date`
+- `dry_run`
+
+Before live stable:
+
+1. resolve the target stable version with `./scripts/release.sh stable --date YYYY-MM-DD --print-version`
+2. ensure `releases/vYYYY.MDD.P.md` exists on the source ref
+3. run the stable workflow in dry-run mode first when practical
+4. then run the real stable publish
+
+The stable workflow:
+
+- re-verifies the exact source ref
+- computes the next stable patch slot for the chosen UTC date
+- publishes `YYYY.MDD.P` under dist-tag `latest`
+- creates git tag `vYYYY.MDD.P`
+- creates or updates the GitHub Release from `releases/vYYYY.MDD.P.md`
+
+Local emergency/manual commands:
+
+```bash
+./scripts/release.sh stable --dry-run
+./scripts/release.sh stable
+git push public-gh refs/tags/vYYYY.MDD.P
+./scripts/create-github-release.sh YYYY.MDD.P
+```
+
+## Step 7 — Finish the Other Surfaces
+
+Create or verify follow-up work for:
+
+- website changelog publishing
+- launch post / social announcement
+- release summary in Paperclip issue context
+
+These should reference the stable release, not the canary.
+
+## Failure Handling
+
+If the canary is bad:
+
+- publish another canary, do not ship stable
+
+If stable npm publish succeeds but tag push or GitHub release creation fails:
+
+- fix the git/GitHub issue immediately from the same release result
+- do not republish the same version
+
+If `latest` is bad after stable publish:
+
+```bash
+./scripts/rollback-latest.sh <last-good-version>
+```
+
+Then fix forward with a new stable release.
+
+## Output
+
+When the skill completes, provide:
+
+- candidate SHA and tested canary version, if relevant
+- stable version, if promoted
+- verification status
+- npm status
+- smoke-test status
+- git tag / GitHub Release status
+- website / announcement follow-up status
+- rollback recommendation if anything is still partially complete

.changeset/README.md

@@ -1,8 +0,0 @@
-# Changesets
-
-Hello and welcome! This folder has been automatically generated by `@changesets/cli`, a build tool that works
-with multi-package repos, or single-package repos to help you version and publish your code. You can
-find the full documentation for it [in our repository](https://github.com/changesets/changesets).
-
-We have a quick list of common questions to get you started engaging with this project in
-[our documentation](https://github.com/changesets/changesets/blob/main/docs/common-questions.md).

.changeset/config.json

@@ -1,11 +0,0 @@
-{
-  "$schema": "https://unpkg.com/@changesets/config@3.1.3/schema.json",
-  "changelog": "@changesets/cli/changelog",
-  "commit": false,
-  "fixed": [["@paperclipai/*", "paperclipai"]],
-  "linked": [],
-  "access": "public",
-  "baseBranch": "master",
-  "updateInternalDependencies": "patch",
-  "ignore": ["@paperclipai/ui"]
-}

.claude/skills/company-creator

@@ -0,0 +1 @@
+../../.agents/skills/company-creator

.github/CODEOWNERS

@@ -0,0 +1,10 @@
+# Replace @cryppadotta if a different maintainer or team should own release infrastructure.
+
+.github/** @cryppadotta @devinfoley
+scripts/release*.sh @cryppadotta @devinfoley
+scripts/release-*.mjs @cryppadotta @devinfoley
+scripts/create-github-release.sh @cryppadotta @devinfoley
+scripts/rollback-latest.sh @cryppadotta @devinfoley
+doc/RELEASING.md @cryppadotta @devinfoley
+doc/PUBLISHING.md @cryppadotta @devinfoley
+doc/RELEASE-AUTOMATION-SETUP.md @cryppadotta @devinfoley

.github/workflows/e2e.yml

@@ -0,0 +1,44 @@
+name: E2E Tests
+
+on:
+  workflow_dispatch:
+    inputs:
+      skip_llm:
+        description: "Skip LLM-dependent assertions (default: true)"
+        type: boolean
+        default: true
+
+jobs:
+  e2e:
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    env:
+      PAPERCLIP_E2E_SKIP_LLM: ${{ inputs.skip_llm && 'true' || 'false' }}
+      ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: pnpm/action-setup@v4
+        with:
+          version: 9
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+          cache: pnpm
+
+      - run: pnpm install --frozen-lockfile
+      - run: pnpm build
+      - run: npx playwright install --with-deps chromium
+
+      - name: Run e2e tests
+        run: pnpm run test:e2e
+
+      - uses: actions/upload-artifact@v4
+        if: always()
+        with:
+          name: playwright-report
+          path: |
+            tests/e2e/playwright-report/
+            tests/e2e/test-results/
+          retention-days: 14

.github/workflows/pr.yml

@@ -0,0 +1,146 @@
+name: PR
+
+on:
+  pull_request:
+    branches:
+      - master
+
+concurrency:
+  group: pr-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+
+jobs:
+  policy:
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Block manual lockfile edits
+        if: github.head_ref != 'chore/refresh-lockfile'
+        run: |
+          changed="$(git diff --name-only "${{ github.event.pull_request.base.sha }}" "${{ github.event.pull_request.head.sha }}")"
+          if printf '%s\n' "$changed" | grep -qx 'pnpm-lock.yaml'; then
+            echo "Do not commit pnpm-lock.yaml in pull requests. CI owns lockfile updates."
+            exit 1
+          fi
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+          run_install: false
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+
+      - name: Validate dependency resolution when manifests change
+        run: |
+          changed="$(git diff --name-only "${{ github.event.pull_request.base.sha }}" "${{ github.event.pull_request.head.sha }}")"
+          manifest_pattern='(^|/)package\.json$|^pnpm-workspace\.yaml$|^\.npmrc$|^pnpmfile\.(cjs|js|mjs)$'
+          if printf '%s\n' "$changed" | grep -Eq "$manifest_pattern"; then
+            pnpm install --lockfile-only --ignore-scripts --no-frozen-lockfile
+          fi
+
+  verify:
+    needs: [policy]
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Typecheck
+        run: pnpm -r typecheck
+
+      - name: Run tests
+        run: pnpm test:run
+
+      - name: Build
+        run: pnpm build
+
+      - name: Release canary dry run
+        run: |
+          git checkout -B master HEAD
+          git checkout -- pnpm-lock.yaml
+          ./scripts/release.sh canary --skip-verify --dry-run
+
+  e2e:
+    needs: [policy]
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Build
+        run: pnpm build
+
+      - name: Install Playwright
+        run: npx playwright install --with-deps chromium
+
+      - name: Generate Paperclip config
+        run: |
+          mkdir -p ~/.paperclip/instances/default
+          cat > ~/.paperclip/instances/default/config.json << 'CONF'
+          {
+            "$meta": { "version": 1, "updatedAt": "2026-01-01T00:00:00.000Z", "source": "onboard" },
+            "database": { "mode": "embedded-postgres" },
+            "logging": { "mode": "file" },
+            "server": { "deploymentMode": "local_trusted", "host": "127.0.0.1", "port": 3100 },
+            "auth": { "baseUrlMode": "auto" },
+            "storage": { "provider": "local_disk" },
+            "secrets": { "provider": "local_encrypted", "strictMode": false }
+          }
+          CONF
+
+      - name: Run e2e tests
+        env:
+          PAPERCLIP_E2E_SKIP_LLM: "true"
+        run: pnpm run test:e2e
+
+      - name: Upload Playwright report
+        uses: actions/upload-artifact@v4
+        if: always()
+        with:
+          name: playwright-report
+          path: |
+            tests/e2e/playwright-report/
+            tests/e2e/test-results/
+          retention-days: 14

.github/workflows/refresh-lockfile.yml

@@ -0,0 +1,97 @@
+name: Refresh Lockfile
+
+on:
+  push:
+    branches:
+      - master
+  workflow_dispatch:
+
+concurrency:
+  group: refresh-lockfile-master
+  cancel-in-progress: false
+
+jobs:
+  refresh:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    permissions:
+      contents: write
+      pull-requests: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+          run_install: false
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 20
+          cache: pnpm
+
+      - name: Refresh pnpm lockfile
+        run: pnpm install --lockfile-only --ignore-scripts --no-frozen-lockfile
+
+      - name: Fail on unexpected file changes
+        run: |
+          changed="$(git status --porcelain)"
+          if [ -z "$changed" ]; then
+            echo "Lockfile is already up to date."
+            exit 0
+          fi
+          if printf '%s\n' "$changed" | grep -Fvq ' pnpm-lock.yaml'; then
+            echo "Unexpected files changed during lockfile refresh:"
+            echo "$changed"
+            exit 1
+          fi
+
+      - name: Create or update pull request
+        id: upsert-pr
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          if git diff --quiet -- pnpm-lock.yaml; then
+            echo "Lockfile unchanged, nothing to do."
+            echo "pr_created=false" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+
+          BRANCH="chore/refresh-lockfile"
+          git config user.name "lockfile-bot"
+          git config user.email "lockfile-bot@users.noreply.github.com"
+
+          git checkout -B "$BRANCH"
+          git add pnpm-lock.yaml
+          git commit -m "chore(lockfile): refresh pnpm-lock.yaml"
+          git push --force origin "$BRANCH"
+
+          # Create PR if one doesn't already exist
+          existing=$(gh pr list --head "$BRANCH" --json number --jq '.[0].number')
+          if [ -z "$existing" ]; then
+            gh pr create \
+              --head "$BRANCH" \
+              --title "chore(lockfile): refresh pnpm-lock.yaml" \
+              --body "Auto-generated lockfile refresh after dependencies changed on master. This PR only updates pnpm-lock.yaml."
+            echo "Created new PR."
+          else
+            echo "PR #$existing already exists, branch updated via force push."
+          fi
+          echo "pr_created=true" >> "$GITHUB_OUTPUT"
+
+      - name: Enable auto-merge for lockfile PR
+        if: steps.upsert-pr.outputs.pr_created == 'true'
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          pr_url="$(gh pr list --head chore/refresh-lockfile --json url --jq '.[0].url')"
+          if [ -z "$pr_url" ]; then
+            echo "Error: lockfile PR was not found." >&2
+            exit 1
+          fi
+
+          gh pr merge --auto --squash --delete-branch "$pr_url"

.github/workflows/release-smoke.yml

@@ -0,0 +1,118 @@
+name: Release Smoke
+
+on:
+  workflow_dispatch:
+    inputs:
+      paperclip_version:
+        description: Published Paperclip dist-tag to test
+        required: true
+        default: canary
+        type: choice
+        options:
+          - canary
+          - latest
+      host_port:
+        description: Host port for the Docker smoke container
+        required: false
+        default: "3232"
+        type: string
+      artifact_name:
+        description: Artifact name for uploaded diagnostics
+        required: false
+        default: release-smoke
+        type: string
+  workflow_call:
+    inputs:
+      paperclip_version:
+        required: true
+        type: string
+      host_port:
+        required: false
+        default: "3232"
+        type: string
+      artifact_name:
+        required: false
+        default: release-smoke
+        type: string
+
+jobs:
+  smoke:
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --no-frozen-lockfile
+
+      - name: Install Playwright browser
+        run: npx playwright install --with-deps chromium
+
+      - name: Launch Docker smoke harness
+        run: |
+          metadata_file="$RUNNER_TEMP/release-smoke.env"
+          HOST_PORT="${{ inputs.host_port }}" \
+          DATA_DIR="$RUNNER_TEMP/release-smoke-data" \
+          PAPERCLIPAI_VERSION="${{ inputs.paperclip_version }}" \
+          SMOKE_DETACH=true \
+          SMOKE_METADATA_FILE="$metadata_file" \
+          ./scripts/docker-onboard-smoke.sh
+          set -a
+          source "$metadata_file"
+          set +a
+          {
+            echo "SMOKE_BASE_URL=$SMOKE_BASE_URL"
+            echo "SMOKE_ADMIN_EMAIL=$SMOKE_ADMIN_EMAIL"
+            echo "SMOKE_ADMIN_PASSWORD=$SMOKE_ADMIN_PASSWORD"
+            echo "SMOKE_CONTAINER_NAME=$SMOKE_CONTAINER_NAME"
+            echo "SMOKE_DATA_DIR=$SMOKE_DATA_DIR"
+            echo "SMOKE_IMAGE_NAME=$SMOKE_IMAGE_NAME"
+            echo "SMOKE_PAPERCLIPAI_VERSION=$SMOKE_PAPERCLIPAI_VERSION"
+            echo "SMOKE_METADATA_FILE=$metadata_file"
+          } >> "$GITHUB_ENV"
+
+      - name: Run release smoke Playwright suite
+        env:
+          PAPERCLIP_RELEASE_SMOKE_BASE_URL: ${{ env.SMOKE_BASE_URL }}
+          PAPERCLIP_RELEASE_SMOKE_EMAIL: ${{ env.SMOKE_ADMIN_EMAIL }}
+          PAPERCLIP_RELEASE_SMOKE_PASSWORD: ${{ env.SMOKE_ADMIN_PASSWORD }}
+        run: pnpm run test:release-smoke
+
+      - name: Capture Docker logs
+        if: always()
+        run: |
+          if [[ -n "${SMOKE_CONTAINER_NAME:-}" ]]; then
+            docker logs "$SMOKE_CONTAINER_NAME" >"$RUNNER_TEMP/docker-onboard-smoke.log" 2>&1 || true
+          fi
+
+      - name: Upload diagnostics
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: ${{ inputs.artifact_name }}
+          path: |
+            ${{ runner.temp }}/docker-onboard-smoke.log
+            ${{ env.SMOKE_METADATA_FILE }}
+            tests/release-smoke/playwright-report/
+            tests/release-smoke/test-results/
+          retention-days: 14
+
+      - name: Stop Docker smoke container
+        if: always()
+        run: |
+          if [[ -n "${SMOKE_CONTAINER_NAME:-}" ]]; then
+            docker rm -f "$SMOKE_CONTAINER_NAME" >/dev/null 2>&1 || true
+          fi

.github/workflows/release.yml

@@ -0,0 +1,261 @@
+name: Release
+
+on:
+  push:
+    branches:
+      - master
+  workflow_dispatch:
+    inputs:
+      source_ref:
+        description: Commit SHA, branch, or tag to publish as stable
+        required: true
+        type: string
+        default: master
+      stable_date:
+        description: Enter a UTC date in YYYY-MM-DD format, for example 2026-03-18. Do not enter a version string. The workflow will resolve that date to a stable version such as 2026.318.0, then 2026.318.1 for the next same-day stable.
+        required: false
+        type: string
+      dry_run:
+        description: Preview the stable release without publishing
+        required: true
+        type: boolean
+        default: false
+
+concurrency:
+  group: release-${{ github.event_name }}-${{ github.ref }}
+  cancel-in-progress: false
+
+jobs:
+  verify_canary:
+    if: github.event_name == 'push'
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    permissions:
+      contents: read
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --no-frozen-lockfile
+
+      - name: Typecheck
+        run: pnpm -r typecheck
+
+      - name: Run tests
+        run: pnpm test:run
+
+      - name: Build
+        run: pnpm build
+
+  publish_canary:
+    if: github.event_name == 'push'
+    needs: verify_canary
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+    environment: npm-canary
+    permissions:
+      contents: write
+      id-token: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --no-frozen-lockfile
+
+      - name: Restore tracked install-time changes
+        run: git checkout -- pnpm-lock.yaml
+
+      - name: Configure git author
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+
+      - name: Publish canary
+        env:
+          GITHUB_ACTIONS: "true"
+        run: ./scripts/release.sh canary --skip-verify
+
+      - name: Push canary tag
+        run: |
+          tag="$(git tag --points-at HEAD | grep '^canary/v' | head -1)"
+          if [ -z "$tag" ]; then
+            echo "Error: no canary tag points at HEAD after release." >&2
+            exit 1
+          fi
+          git push origin "refs/tags/${tag}"
+
+  verify_stable:
+    if: github.event_name == 'workflow_dispatch'
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    permissions:
+      contents: read
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: ${{ inputs.source_ref }}
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --no-frozen-lockfile
+
+      - name: Typecheck
+        run: pnpm -r typecheck
+
+      - name: Run tests
+        run: pnpm test:run
+
+      - name: Build
+        run: pnpm build
+
+  preview_stable:
+    if: github.event_name == 'workflow_dispatch' && inputs.dry_run
+    needs: verify_stable
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+    permissions:
+      contents: read
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: ${{ inputs.source_ref }}
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --no-frozen-lockfile
+
+      - name: Dry-run stable release
+        env:
+          GITHUB_ACTIONS: "true"
+        run: |
+          args=(stable --skip-verify --dry-run)
+          if [ -n "${{ inputs.stable_date }}" ]; then
+            args+=(--date "${{ inputs.stable_date }}")
+          fi
+          ./scripts/release.sh "${args[@]}"
+
+  publish_stable:
+    if: github.event_name == 'workflow_dispatch' && !inputs.dry_run
+    needs: verify_stable
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+    environment: npm-stable
+    permissions:
+      contents: write
+      id-token: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: ${{ inputs.source_ref }}
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 9.15.4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --no-frozen-lockfile
+
+      - name: Restore tracked install-time changes
+        run: git checkout -- pnpm-lock.yaml
+
+      - name: Configure git author
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+
+      - name: Publish stable
+        env:
+          GITHUB_ACTIONS: "true"
+        run: |
+          args=(stable --skip-verify)
+          if [ -n "${{ inputs.stable_date }}" ]; then
+            args+=(--date "${{ inputs.stable_date }}")
+          fi
+          ./scripts/release.sh "${args[@]}"
+
+      - name: Push stable tag
+        run: |
+          tag="$(git tag --points-at HEAD | grep '^v' | head -1)"
+          if [ -z "$tag" ]; then
+            echo "Error: no stable tag points at HEAD after release." >&2
+            exit 1
+          fi
+          git push origin "refs/tags/${tag}"
+
+      - name: Create GitHub Release
+        env:
+          GH_TOKEN: ${{ github.token }}
+          PUBLISH_REMOTE: origin
+        run: |
+          version="$(git tag --points-at HEAD | grep '^v' | head -1 | sed 's/^v//')"
+          if [ -z "$version" ]; then
+            echo "Error: no v* tag points at HEAD after stable release." >&2
+            exit 1
+          fi
+          ./scripts/create-github-release.sh "$version"

.gitignore

@@ -36,4 +36,17 @@ tmp/
 *.tmp
 .vscode/
 .claude/settings.local.json
-.paperclip-local/
+.paperclip-local/
+/.idea/
+/.agents/
+
+# Doc maintenance cursor
+.doc-review-cursor
+
+# Playwright
+tests/e2e/test-results/
+tests/e2e/playwright-report/
+tests/release-smoke/test-results/
+tests/release-smoke/playwright-report/
+.superset/
+.claude/worktrees/

AGENTS.md

@@ -78,6 +78,9 @@ If you change schema/API behavior, update all impacted layers:
 4. Do not replace strategic docs wholesale unless asked.
 Prefer additive updates. Keep `doc/SPEC.md` and `doc/SPEC-implementation.md` aligned.
 
+5. Keep plan docs dated and centralized.
+New plan documents belong in `doc/plans/` and should use `YYYY-MM-DD-slug.md` filenames.
+
 ## 6. Database Change Workflow
 
 When changing data model:

CONTRIBUTING.md

@@ -0,0 +1,74 @@
+# Contributing Guide
+
+Thanks for wanting to contribute!
+
+We really appreciate both small fixes and thoughtful larger changes.
+
+## Two Paths to Get Your Pull Request Accepted
+
+### Path 1: Small, Focused Changes (Fastest way to get merged)
+
+- Pick **one** clear thing to fix/improve
+- Touch the **smallest possible number of files**
+- Make sure the change is very targeted and easy to review
+- All automated checks pass (including Greptile comments)
+- No new lint/test failures
+
+These almost always get merged quickly when they're clean.
+
+### Path 2: Bigger or Impactful Changes
+
+- **First** talk about it in Discord → #dev channel  
+  → Describe what you're trying to solve  
+  → Share rough ideas / approach
+- Once there's rough agreement, build it
+- In your PR include:
+  - Before / After screenshots (or short video if UI/behavior change)
+  - Clear description of what & why
+  - Proof it works (manual testing notes)
+  - All tests passing
+  - All Greptile + other PR comments addressed
+
+PRs that follow this path are **much** more likely to be accepted, even when they're large.
+
+## General Rules (both paths)
+
+- Write clear commit messages
+- Keep PR title + description meaningful
+- One PR = one logical change (unless it's a small related group)
+- Run tests locally first
+- Be kind in discussions 😄
+
+## Writing a Good PR message
+
+Please include a "thinking path" at the top of your PR message that explains from the top of the project down to what you fixed. E.g.:
+
+### Thinking Path Example 1:
+
+> - Paperclip orchestrates ai-agents for zero-human companies
+> - There are many types of adapters for each LLM model provider
+> - But LLM's have a context limit and not all agents can automatically compact their context
+> - So we need to have an adapter-specific configuration for which adapters can and cannot automatically compact their context
+> - This pull request adds per-adapter configuration of compaction, either auto or paperclip managed
+> - That way we can get optimal performance from any adapter/provider in Paperclip
+
+### Thinking Path Example 2:
+
+> - Paperclip orchestrates ai-agents for zero-human companies
+> - But humans want to watch the agents and oversee their work
+> - Human users also operate in teams and so they need their own logins, profiles, views etc.
+> - So we have a multi-user system for humans
+> - But humans want to be able to update their own profile picture and avatar
+> - But the avatar upload form wasn't saving the avatar to the file storage system
+> - So this PR fixes the avatar upload form to use the file storage service
+> - The benefit is we don't have a one-off file storage for just one aspect of the system, which would cause confusion and extra configuration
+
+Then have the rest of your normal PR message after the Thinking Path.
+
+This should include details about what you did, why you did it, why it matters & the benefits, how we can verify it works, and any risks.
+
+Please include screenshots if possible if you have a visible change. (use something like the [agent-browser skill](https://github.com/vercel-labs/agent-browser/blob/main/skills/agent-browser/SKILL.md) or similar to take screenshots). Ideally, you include before and after screenshots.
+
+Questions? Just ask in #dev — we're happy to help.
+
+Happy hacking!

Dockerfile

@@ -1,4 +1,4 @@
-FROM node:20-bookworm-slim AS base
+FROM node:lts-trixie-slim AS base
 RUN apt-get update \
   && apt-get install -y --no-install-recommends ca-certificates curl git \
   && rm -rf /var/lib/apt/lists/*
@@ -15,19 +15,28 @@ COPY packages/db/package.json packages/db/
 COPY packages/adapter-utils/package.json packages/adapter-utils/
 COPY packages/adapters/claude-local/package.json packages/adapters/claude-local/
 COPY packages/adapters/codex-local/package.json packages/adapters/codex-local/
+COPY packages/adapters/cursor-local/package.json packages/adapters/cursor-local/
+COPY packages/adapters/gemini-local/package.json packages/adapters/gemini-local/
+COPY packages/adapters/openclaw-gateway/package.json packages/adapters/openclaw-gateway/
+COPY packages/adapters/opencode-local/package.json packages/adapters/opencode-local/
+COPY packages/adapters/pi-local/package.json packages/adapters/pi-local/
+
 RUN pnpm install --frozen-lockfile
 
 FROM base AS build
 WORKDIR /app
 COPY --from=deps /app /app
 COPY . .
-RUN pnpm --filter @paperclip/ui build
-RUN pnpm --filter @paperclip/server build
+RUN pnpm --filter @paperclipai/ui build
+RUN pnpm --filter @paperclipai/server build
+RUN test -f server/dist/index.js || (echo "ERROR: server build output missing" && exit 1)
 
 FROM base AS production
 WORKDIR /app
-COPY --from=build /app /app
-RUN npm install --global --omit=dev @anthropic-ai/claude-code@latest @openai/codex@latest
+COPY --chown=node:node --from=build /app /app
+RUN npm install --global --omit=dev @anthropic-ai/claude-code@latest @openai/codex@latest opencode-ai \
+  && mkdir -p /paperclip \
+  && chown node:node /paperclip
 
 ENV NODE_ENV=production \
   HOME=/paperclip \
@@ -37,10 +46,11 @@ ENV NODE_ENV=production \
   PAPERCLIP_HOME=/paperclip \
   PAPERCLIP_INSTANCE_ID=default \
   PAPERCLIP_CONFIG=/paperclip/instances/default/config.json \
-  PAPERCLIP_DEPLOYMENT_MODE=local_trusted \
+  PAPERCLIP_DEPLOYMENT_MODE=authenticated \
   PAPERCLIP_DEPLOYMENT_EXPOSURE=private
 
 VOLUME ["/paperclip"]
 EXPOSE 3100
 
+USER node
 CMD ["node", "--import", "./server/node_modules/tsx/dist/loader.mjs", "server/dist/index.js"]

Dockerfile.onboard-smoke

@@ -2,17 +2,22 @@ FROM ubuntu:24.04
 
 ARG NODE_MAJOR=20
 ARG PAPERCLIPAI_VERSION=latest
+ARG HOST_UID=10001
 
 ENV DEBIAN_FRONTEND=noninteractive \
   PAPERCLIP_HOME=/paperclip \
   PAPERCLIP_OPEN_ON_LISTEN=false \
   HOST=0.0.0.0 \
   PORT=3100 \
+  HOME=/home/paperclip \
+  LANG=en_US.UTF-8 \
+  LC_ALL=en_US.UTF-8 \
+  NPM_CONFIG_UPDATE_NOTIFIER=false \
   NODE_MAJOR=${NODE_MAJOR} \
   PAPERCLIPAI_VERSION=${PAPERCLIPAI_VERSION}
 
 RUN apt-get update \
-  && apt-get install -y --no-install-recommends ca-certificates curl gnupg \
+  && apt-get install -y --no-install-recommends ca-certificates curl gnupg locales \
   && mkdir -p /etc/apt/keyrings \
   && curl -fsSL https://deb.nodesource.com/gpgkey/nodesource-repo.gpg.key \
     | gpg --dearmor -o /etc/apt/keyrings/nodesource.gpg \
@@ -20,10 +25,16 @@ RUN apt-get update \
     > /etc/apt/sources.list.d/nodesource.list \
   && apt-get update \
   && apt-get install -y --no-install-recommends nodejs \
+  && locale-gen en_US.UTF-8 \
+  && groupadd --gid 10001 paperclip \
+  && useradd --create-home --shell /bin/bash --uid "${HOST_UID}" --gid 10001 paperclip \
+  && mkdir -p /paperclip /home/paperclip/workspace \
+  && chown -R paperclip:paperclip /paperclip /home/paperclip \
   && rm -rf /var/lib/apt/lists/*
 
 VOLUME ["/paperclip"]
-WORKDIR /workspace
+WORKDIR /home/paperclip/workspace
 EXPOSE 3100
+USER paperclip
 
 CMD ["bash", "-lc", "set -euo pipefail; mkdir -p \"$PAPERCLIP_HOME\"; npx --yes \"paperclipai@${PAPERCLIPAI_VERSION}\" onboard --yes --data-dir \"$PAPERCLIP_HOME\""]

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Paperclip AI
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -218,7 +218,8 @@ By default, agents run on scheduled heartbeats and event-based triggers (task as
 ## Development
 
 ```bash
-pnpm dev              # Full dev (API + UI)
+pnpm dev              # Full dev (API + UI, watch mode)
+pnpm dev:once         # Full dev without file watching
 pnpm dev:server       # Server only
 pnpm build            # Build all
 pnpm typecheck        # Type checking
@@ -233,9 +234,13 @@ See [doc/DEVELOPING.md](doc/DEVELOPING.md) for the full development guide.
 
 ## Roadmap
 
-- 🛒 **Clipmart** — Download and share entire company architectures
-- 🧩 **Plugin System** — Embed custom plugins (e.g. Reporting, Knowledge Base) into Paperclip
-- ☁️ **Cloud Agent Adapters** — Add more adapters for cloud-hosted agents
+- ⚪ Get OpenClaw onboarding easier
+- ⚪ Get cloud agents working e.g. Cursor / e2b agents
+- ⚪ ClipMart - buy and sell entire agent companies
+- ⚪ Easy agent configurations / easier to understand
+- ⚪ Better support for harness engineering
+- 🟢 Plugin system (e.g. if you want to add a knowledgebase, custom tracing, queues, etc)
+- ⚪ Better docs
 
 <br/>
 
@@ -243,8 +248,6 @@ See [doc/DEVELOPING.md](doc/DEVELOPING.md) for the full development guide.
 
 We welcome contributions. See the [contributing guide](CONTRIBUTING.md) for details.
 
-<!-- TODO: add CONTRIBUTING.md -->
-
 <br/>
 
 ## Community
@@ -259,6 +262,10 @@ We welcome contributions. See the [contributing guide](CONTRIBUTING.md) for deta
 
 MIT &copy; 2026 Paperclip
 
+## Star History
+
+[![Star History Chart](https://api.star-history.com/image?repos=paperclipai/paperclip&type=date&legend=top-left)](https://www.star-history.com/?repos=paperclipai%2Fpaperclip&type=date&legend=top-left)
+
 <br/>
 
 ---

cli/CHANGELOG.md

@@ -1,5 +1,72 @@
 # paperclipai
 
+## 0.3.1
+
+### Patch Changes
+
+- Stable release preparation for 0.3.1
+- Updated dependencies
+  - @paperclipai/adapter-utils@0.3.1
+  - @paperclipai/adapter-claude-local@0.3.1
+  - @paperclipai/adapter-codex-local@0.3.1
+  - @paperclipai/adapter-cursor-local@0.3.1
+  - @paperclipai/adapter-gemini-local@0.3.1
+  - @paperclipai/adapter-openclaw-gateway@0.3.1
+  - @paperclipai/adapter-opencode-local@0.3.1
+  - @paperclipai/adapter-pi-local@0.3.1
+  - @paperclipai/db@0.3.1
+  - @paperclipai/shared@0.3.1
+  - @paperclipai/server@0.3.1
+
+## 0.3.0
+
+### Minor Changes
+
+- Stable release preparation for 0.3.0
+
+### Patch Changes
+
+- Updated dependencies [6077ae6]
+- Updated dependencies
+  - @paperclipai/shared@0.3.0
+  - @paperclipai/adapter-utils@0.3.0
+  - @paperclipai/adapter-claude-local@0.3.0
+  - @paperclipai/adapter-codex-local@0.3.0
+  - @paperclipai/adapter-cursor-local@0.3.0
+  - @paperclipai/adapter-openclaw-gateway@0.3.0
+  - @paperclipai/adapter-opencode-local@0.3.0
+  - @paperclipai/adapter-pi-local@0.3.0
+  - @paperclipai/db@0.3.0
+  - @paperclipai/server@0.3.0
+
+## 0.2.7
+
+### Patch Changes
+
+- Version bump (patch)
+- Updated dependencies
+  - @paperclipai/shared@0.2.7
+  - @paperclipai/adapter-utils@0.2.7
+  - @paperclipai/db@0.2.7
+  - @paperclipai/adapter-claude-local@0.2.7
+  - @paperclipai/adapter-codex-local@0.2.7
+  - @paperclipai/adapter-openclaw@0.2.7
+  - @paperclipai/server@0.2.7
+
+## 0.2.6
+
+### Patch Changes
+
+- Version bump (patch)
+- Updated dependencies
+  - @paperclipai/shared@0.2.6
+  - @paperclipai/adapter-utils@0.2.6
+  - @paperclipai/db@0.2.6
+  - @paperclipai/adapter-claude-local@0.2.6
+  - @paperclipai/adapter-codex-local@0.2.6
+  - @paperclipai/adapter-openclaw@0.2.6
+  - @paperclipai/server@0.2.6
+
 ## 0.2.5
 
 ### Patch Changes

cli/esbuild.config.mjs

@@ -21,7 +21,7 @@ const workspacePaths = [
   "packages/adapter-utils",
   "packages/adapters/claude-local",
   "packages/adapters/codex-local",
-  "packages/adapters/openclaw",
+  "packages/adapters/openclaw-gateway",
 ];
 
 // Workspace packages that should NOT be bundled — they'll be published

cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "paperclipai",
-  "version": "0.2.5",
+  "version": "0.3.1",
   "description": "Paperclip CLI — orchestrate AI agent teams to run a business",
   "type": "module",
   "bin": {
@@ -16,10 +16,13 @@
   "license": "MIT",
   "repository": {
     "type": "git",
-    "url": "https://github.com/paperclipai/paperclip.git",
+    "url": "https://github.com/paperclipai/paperclip",
     "directory": "cli"
   },
   "homepage": "https://github.com/paperclipai/paperclip",
+  "bugs": {
+    "url": "https://github.com/paperclipai/paperclip/issues"
+  },
   "files": [
     "dist"
   ],
@@ -36,7 +39,11 @@
     "@clack/prompts": "^0.10.0",
     "@paperclipai/adapter-claude-local": "workspace:*",
     "@paperclipai/adapter-codex-local": "workspace:*",
-    "@paperclipai/adapter-openclaw": "workspace:*",
+    "@paperclipai/adapter-cursor-local": "workspace:*",
+    "@paperclipai/adapter-gemini-local": "workspace:*",
+    "@paperclipai/adapter-opencode-local": "workspace:*",
+    "@paperclipai/adapter-pi-local": "workspace:*",
+    "@paperclipai/adapter-openclaw-gateway": "workspace:*",
     "@paperclipai/adapter-utils": "workspace:*",
     "@paperclipai/db": "workspace:*",
     "@paperclipai/server": "workspace:*",
@@ -44,6 +51,7 @@
     "drizzle-orm": "0.38.4",
     "dotenv": "^17.0.1",
     "commander": "^13.1.0",
+    "embedded-postgres": "^18.1.0-beta.16",
     "picocolors": "^1.1.1"
   },
   "devDependencies": {

cli/src/__tests__/agent-jwt-env.test.ts

@@ -4,7 +4,9 @@ import path from "node:path";
 import { afterEach, beforeEach, describe, expect, it } from "vitest";
 import {
   ensureAgentJwtSecret,
+  mergePaperclipEnvEntries,
   readAgentJwtSecretFromEnv,
+  readPaperclipEnvEntries,
   resolveAgentJwtEnvFile,
 } from "../config/env.js";
 import { agentJwtSecretCheck } from "../checks/agent-jwt-secret-check.js";
@@ -58,4 +60,20 @@ describe("agent jwt env helpers", () => {
     const result = agentJwtSecretCheck(configPath);
     expect(result.status).toBe("pass");
   });
+
+  it("quotes hash-prefixed env values so dotenv round-trips them", () => {
+    const configPath = tempConfigPath();
+    const envPath = resolveAgentJwtEnvFile(configPath);
+
+    mergePaperclipEnvEntries(
+      {
+        PAPERCLIP_WORKTREE_COLOR: "#439edb",
+      },
+      envPath,
+    );
+
+    const contents = fs.readFileSync(envPath, "utf-8");
+    expect(contents).toContain('PAPERCLIP_WORKTREE_COLOR="#439edb"');
+    expect(readPaperclipEnvEntries(envPath).PAPERCLIP_WORKTREE_COLOR).toBe("#439edb");
+  });
 });

cli/src/__tests__/allowed-hostname.test.ts

@@ -21,6 +21,12 @@ function writeBaseConfig(configPath: string) {
       mode: "embedded-postgres",
       embeddedPostgresDataDir: "/tmp/paperclip-db",
       embeddedPostgresPort: 54329,
+      backup: {
+        enabled: true,
+        intervalMinutes: 60,
+        retentionDays: 30,
+        dir: "/tmp/paperclip-backups",
+      },
     },
     logging: {
       mode: "file",
@@ -36,6 +42,7 @@ function writeBaseConfig(configPath: string) {
     },
     auth: {
       baseUrlMode: "auto",
+      disableSignUp: false,
     },
     storage: {
       provider: "local_disk",
@@ -68,4 +75,3 @@ describe("allowed-hostname command", () => {
     expect(raw.server.allowedHostnames).toEqual(["dotta-macbook-pro"]);
   });
 });
-

cli/src/__tests__/company-delete.test.ts

@@ -8,12 +8,16 @@ function makeCompany(overrides: Partial<Company>): Company {
     name: "Alpha",
     description: null,
     status: "active",
+    pauseReason: null,
+    pausedAt: null,
     issuePrefix: "ALP",
     issueCounter: 1,
     budgetMonthlyCents: 0,
     spentMonthlyCents: 0,
     requireBoardApprovalForNewAgents: false,
     brandColor: null,
+    logoAssetId: null,
+    logoUrl: null,
     createdAt: new Date(),
     updatedAt: new Date(),
     ...overrides,

cli/src/__tests__/company-import-url.test.ts

@@ -0,0 +1,31 @@
+import { describe, expect, it } from "vitest";
+import { isHttpUrl, isGithubUrl } from "../commands/client/company.js";
+
+describe("isHttpUrl", () => {
+  it("matches http URLs", () => {
+    expect(isHttpUrl("http://example.com/foo")).toBe(true);
+  });
+
+  it("matches https URLs", () => {
+    expect(isHttpUrl("https://example.com/foo")).toBe(true);
+  });
+
+  it("rejects local paths", () => {
+    expect(isHttpUrl("/tmp/my-company")).toBe(false);
+    expect(isHttpUrl("./relative")).toBe(false);
+  });
+});
+
+describe("isGithubUrl", () => {
+  it("matches GitHub URLs", () => {
+    expect(isGithubUrl("https://github.com/org/repo")).toBe(true);
+  });
+
+  it("rejects non-GitHub HTTP URLs", () => {
+    expect(isGithubUrl("https://example.com/foo")).toBe(false);
+  });
+
+  it("rejects local paths", () => {
+    expect(isGithubUrl("/tmp/my-company")).toBe(false);
+  });
+});

cli/src/__tests__/doctor.test.ts

@@ -0,0 +1,99 @@
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import { afterEach, beforeEach, describe, expect, it } from "vitest";
+import { doctor } from "../commands/doctor.js";
+import { writeConfig } from "../config/store.js";
+import type { PaperclipConfig } from "../config/schema.js";
+
+const ORIGINAL_ENV = { ...process.env };
+
+function createTempConfig(): string {
+  const root = fs.mkdtempSync(path.join(os.tmpdir(), "paperclip-doctor-"));
+  const configPath = path.join(root, ".paperclip", "config.json");
+  const runtimeRoot = path.join(root, "runtime");
+
+  const config: PaperclipConfig = {
+    $meta: {
+      version: 1,
+      updatedAt: "2026-03-10T00:00:00.000Z",
+      source: "configure",
+    },
+    database: {
+      mode: "embedded-postgres",
+      embeddedPostgresDataDir: path.join(runtimeRoot, "db"),
+      embeddedPostgresPort: 55432,
+      backup: {
+        enabled: true,
+        intervalMinutes: 60,
+        retentionDays: 30,
+        dir: path.join(runtimeRoot, "backups"),
+      },
+    },
+    logging: {
+      mode: "file",
+      logDir: path.join(runtimeRoot, "logs"),
+    },
+    server: {
+      deploymentMode: "local_trusted",
+      exposure: "private",
+      host: "127.0.0.1",
+      port: 3199,
+      allowedHostnames: [],
+      serveUi: true,
+    },
+    auth: {
+      baseUrlMode: "auto",
+      disableSignUp: false,
+    },
+    storage: {
+      provider: "local_disk",
+      localDisk: {
+        baseDir: path.join(runtimeRoot, "storage"),
+      },
+      s3: {
+        bucket: "paperclip",
+        region: "us-east-1",
+        prefix: "",
+        forcePathStyle: false,
+      },
+    },
+    secrets: {
+      provider: "local_encrypted",
+      strictMode: false,
+      localEncrypted: {
+        keyFilePath: path.join(runtimeRoot, "secrets", "master.key"),
+      },
+    },
+  };
+
+  writeConfig(config, configPath);
+  return configPath;
+}
+
+describe("doctor", () => {
+  beforeEach(() => {
+    process.env = { ...ORIGINAL_ENV };
+    delete process.env.PAPERCLIP_AGENT_JWT_SECRET;
+    delete process.env.PAPERCLIP_SECRETS_MASTER_KEY;
+    delete process.env.PAPERCLIP_SECRETS_MASTER_KEY_FILE;
+  });
+
+  afterEach(() => {
+    process.env = { ...ORIGINAL_ENV };
+  });
+
+  it("re-runs repairable checks so repaired failures do not remain blocking", async () => {
+    const configPath = createTempConfig();
+
+    const summary = await doctor({
+      config: configPath,
+      repair: true,
+      yes: true,
+    });
+
+    expect(summary.failed).toBe(0);
+    expect(summary.warned).toBe(0);
+    expect(process.env.PAPERCLIP_AGENT_JWT_SECRET).toBeTruthy();
+  });
+});

cli/src/__tests__/worktree-merge-history.test.ts

@@ -0,0 +1,392 @@
+import { describe, expect, it } from "vitest";
+import { buildWorktreeMergePlan, parseWorktreeMergeScopes } from "../commands/worktree-merge-history-lib.js";
+
+function makeIssue(overrides: Record<string, unknown> = {}) {
+  return {
+    id: "issue-1",
+    companyId: "company-1",
+    projectId: null,
+    projectWorkspaceId: null,
+    goalId: "goal-1",
+    parentId: null,
+    title: "Issue",
+    description: null,
+    status: "todo",
+    priority: "medium",
+    assigneeAgentId: null,
+    assigneeUserId: null,
+    checkoutRunId: null,
+    executionRunId: null,
+    executionAgentNameKey: null,
+    executionLockedAt: null,
+    createdByAgentId: null,
+    createdByUserId: "local-board",
+    issueNumber: 1,
+    identifier: "PAP-1",
+    requestDepth: 0,
+    billingCode: null,
+    assigneeAdapterOverrides: null,
+    executionWorkspaceId: null,
+    executionWorkspacePreference: null,
+    executionWorkspaceSettings: null,
+    startedAt: null,
+    completedAt: null,
+    cancelledAt: null,
+    hiddenAt: null,
+    createdAt: new Date("2026-03-20T00:00:00.000Z"),
+    updatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    ...overrides,
+  } as any;
+}
+
+function makeComment(overrides: Record<string, unknown> = {}) {
+  return {
+    id: "comment-1",
+    companyId: "company-1",
+    issueId: "issue-1",
+    authorAgentId: null,
+    authorUserId: "local-board",
+    body: "hello",
+    createdAt: new Date("2026-03-20T00:00:00.000Z"),
+    updatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    ...overrides,
+  } as any;
+}
+
+function makeIssueDocument(overrides: Record<string, unknown> = {}) {
+  return {
+    id: "issue-document-1",
+    companyId: "company-1",
+    issueId: "issue-1",
+    documentId: "document-1",
+    key: "plan",
+    linkCreatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    linkUpdatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    title: "Plan",
+    format: "markdown",
+    latestBody: "# Plan",
+    latestRevisionId: "revision-1",
+    latestRevisionNumber: 1,
+    createdByAgentId: null,
+    createdByUserId: "local-board",
+    updatedByAgentId: null,
+    updatedByUserId: "local-board",
+    documentCreatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    documentUpdatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    ...overrides,
+  } as any;
+}
+
+function makeDocumentRevision(overrides: Record<string, unknown> = {}) {
+  return {
+    id: "revision-1",
+    companyId: "company-1",
+    documentId: "document-1",
+    revisionNumber: 1,
+    body: "# Plan",
+    changeSummary: null,
+    createdByAgentId: null,
+    createdByUserId: "local-board",
+    createdAt: new Date("2026-03-20T00:00:00.000Z"),
+    ...overrides,
+  } as any;
+}
+
+function makeAttachment(overrides: Record<string, unknown> = {}) {
+  return {
+    id: "attachment-1",
+    companyId: "company-1",
+    issueId: "issue-1",
+    issueCommentId: null,
+    assetId: "asset-1",
+    provider: "local_disk",
+    objectKey: "company-1/issues/issue-1/2026/03/20/asset.png",
+    contentType: "image/png",
+    byteSize: 12,
+    sha256: "deadbeef",
+    originalFilename: "asset.png",
+    createdByAgentId: null,
+    createdByUserId: "local-board",
+    assetCreatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    assetUpdatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    attachmentCreatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    attachmentUpdatedAt: new Date("2026-03-20T00:00:00.000Z"),
+    ...overrides,
+  } as any;
+}
+
+describe("worktree merge history planner", () => {
+  it("parses default scopes", () => {
+    expect(parseWorktreeMergeScopes(undefined)).toEqual(["issues", "comments"]);
+    expect(parseWorktreeMergeScopes("issues")).toEqual(["issues"]);
+  });
+
+  it("dedupes nested worktree issues by preserved source uuid", () => {
+    const sharedIssue = makeIssue({ id: "issue-a", identifier: "PAP-10", title: "Shared" });
+    const branchOneIssue = makeIssue({
+      id: "issue-b",
+      identifier: "PAP-22",
+      title: "Branch one issue",
+      createdAt: new Date("2026-03-20T01:00:00.000Z"),
+    });
+    const branchTwoIssue = makeIssue({
+      id: "issue-c",
+      identifier: "PAP-23",
+      title: "Branch two issue",
+      createdAt: new Date("2026-03-20T02:00:00.000Z"),
+    });
+
+    const plan = buildWorktreeMergePlan({
+      companyId: "company-1",
+      companyName: "Paperclip",
+      issuePrefix: "PAP",
+      previewIssueCounterStart: 500,
+      scopes: ["issues", "comments"],
+      sourceIssues: [sharedIssue, branchOneIssue, branchTwoIssue],
+      targetIssues: [sharedIssue, branchOneIssue],
+      sourceComments: [],
+      targetComments: [],
+      targetAgents: [],
+      targetProjects: [],
+      targetProjectWorkspaces: [],
+      targetGoals: [{ id: "goal-1" }] as any,
+    });
+
+    expect(plan.counts.issuesToInsert).toBe(1);
+    expect(plan.issuePlans.filter((item) => item.action === "insert").map((item) => item.source.id)).toEqual(["issue-c"]);
+    expect(plan.issuePlans.find((item) => item.source.id === "issue-c" && item.action === "insert")).toMatchObject({
+      previewIdentifier: "PAP-501",
+    });
+  });
+
+  it("clears missing references and coerces in_progress without an assignee", () => {
+    const plan = buildWorktreeMergePlan({
+      companyId: "company-1",
+      companyName: "Paperclip",
+      issuePrefix: "PAP",
+      previewIssueCounterStart: 10,
+      scopes: ["issues"],
+      sourceIssues: [
+        makeIssue({
+          id: "issue-x",
+          identifier: "PAP-99",
+          status: "in_progress",
+          assigneeAgentId: "agent-missing",
+          projectId: "project-missing",
+          projectWorkspaceId: "workspace-missing",
+          goalId: "goal-missing",
+        }),
+      ],
+      targetIssues: [],
+      sourceComments: [],
+      targetComments: [],
+      targetAgents: [],
+      targetProjects: [],
+      targetProjectWorkspaces: [],
+      targetGoals: [],
+    });
+
+    const insert = plan.issuePlans[0] as any;
+    expect(insert.targetStatus).toBe("todo");
+    expect(insert.targetAssigneeAgentId).toBeNull();
+    expect(insert.targetProjectId).toBeNull();
+    expect(insert.targetProjectWorkspaceId).toBeNull();
+    expect(insert.targetGoalId).toBeNull();
+    expect(insert.adjustments).toEqual([
+      "clear_assignee_agent",
+      "clear_project",
+      "clear_project_workspace",
+      "clear_goal",
+      "coerce_in_progress_to_todo",
+    ]);
+  });
+
+  it("applies an explicit project mapping override instead of clearing the project", () => {
+    const plan = buildWorktreeMergePlan({
+      companyId: "company-1",
+      companyName: "Paperclip",
+      issuePrefix: "PAP",
+      previewIssueCounterStart: 10,
+      scopes: ["issues"],
+      sourceIssues: [
+        makeIssue({
+          id: "issue-project-map",
+          identifier: "PAP-77",
+          projectId: "source-project-1",
+          projectWorkspaceId: "source-workspace-1",
+        }),
+      ],
+      targetIssues: [],
+      sourceComments: [],
+      targetComments: [],
+      targetAgents: [],
+      targetProjects: [{ id: "target-project-1", name: "Mapped project", status: "in_progress" }] as any,
+      targetProjectWorkspaces: [],
+      targetGoals: [{ id: "goal-1" }] as any,
+      projectIdOverrides: {
+        "source-project-1": "target-project-1",
+      },
+    });
+
+    const insert = plan.issuePlans[0] as any;
+    expect(insert.targetProjectId).toBe("target-project-1");
+    expect(insert.projectResolution).toBe("mapped");
+    expect(insert.mappedProjectName).toBe("Mapped project");
+    expect(insert.targetProjectWorkspaceId).toBeNull();
+    expect(insert.adjustments).toEqual(["clear_project_workspace"]);
+  });
+
+  it("imports comments onto shared or newly imported issues while skipping existing comments", () => {
+    const sharedIssue = makeIssue({ id: "issue-a", identifier: "PAP-10" });
+    const newIssue = makeIssue({
+      id: "issue-b",
+      identifier: "PAP-11",
+      createdAt: new Date("2026-03-20T01:00:00.000Z"),
+    });
+    const existingComment = makeComment({ id: "comment-existing", issueId: "issue-a" });
+    const sharedIssueComment = makeComment({ id: "comment-shared", issueId: "issue-a" });
+    const newIssueComment = makeComment({
+      id: "comment-new-issue",
+      issueId: "issue-b",
+      authorAgentId: "missing-agent",
+      createdAt: new Date("2026-03-20T01:05:00.000Z"),
+    });
+
+    const plan = buildWorktreeMergePlan({
+      companyId: "company-1",
+      companyName: "Paperclip",
+      issuePrefix: "PAP",
+      previewIssueCounterStart: 10,
+      scopes: ["issues", "comments"],
+      sourceIssues: [sharedIssue, newIssue],
+      targetIssues: [sharedIssue],
+      sourceComments: [existingComment, sharedIssueComment, newIssueComment],
+      targetComments: [existingComment],
+      targetAgents: [],
+      targetProjects: [],
+      targetProjectWorkspaces: [],
+      targetGoals: [{ id: "goal-1" }] as any,
+    });
+
+    expect(plan.counts.commentsToInsert).toBe(2);
+    expect(plan.counts.commentsExisting).toBe(1);
+    expect(plan.commentPlans.filter((item) => item.action === "insert").map((item) => item.source.id)).toEqual([
+      "comment-shared",
+      "comment-new-issue",
+    ]);
+    expect(plan.adjustments.clear_author_agent).toBe(1);
+  });
+
+  it("merges document revisions onto an existing shared document and renumbers conflicts", () => {
+    const sharedIssue = makeIssue({ id: "issue-a", identifier: "PAP-10" });
+    const sourceDocument = makeIssueDocument({
+      issueId: "issue-a",
+      documentId: "document-a",
+      latestBody: "# Branch plan",
+      latestRevisionId: "revision-branch-2",
+      latestRevisionNumber: 2,
+      documentUpdatedAt: new Date("2026-03-20T02:00:00.000Z"),
+      linkUpdatedAt: new Date("2026-03-20T02:00:00.000Z"),
+    });
+    const targetDocument = makeIssueDocument({
+      issueId: "issue-a",
+      documentId: "document-a",
+      latestBody: "# Main plan",
+      latestRevisionId: "revision-main-2",
+      latestRevisionNumber: 2,
+      documentUpdatedAt: new Date("2026-03-20T01:00:00.000Z"),
+      linkUpdatedAt: new Date("2026-03-20T01:00:00.000Z"),
+    });
+    const sourceRevisionOne = makeDocumentRevision({ documentId: "document-a", id: "revision-1" });
+    const sourceRevisionTwo = makeDocumentRevision({
+      documentId: "document-a",
+      id: "revision-branch-2",
+      revisionNumber: 2,
+      body: "# Branch plan",
+      createdAt: new Date("2026-03-20T02:00:00.000Z"),
+    });
+    const targetRevisionOne = makeDocumentRevision({ documentId: "document-a", id: "revision-1" });
+    const targetRevisionTwo = makeDocumentRevision({
+      documentId: "document-a",
+      id: "revision-main-2",
+      revisionNumber: 2,
+      body: "# Main plan",
+      createdAt: new Date("2026-03-20T01:00:00.000Z"),
+    });
+
+    const plan = buildWorktreeMergePlan({
+      companyId: "company-1",
+      companyName: "Paperclip",
+      issuePrefix: "PAP",
+      previewIssueCounterStart: 10,
+      scopes: ["issues", "comments"],
+      sourceIssues: [sharedIssue],
+      targetIssues: [sharedIssue],
+      sourceComments: [],
+      targetComments: [],
+      sourceDocuments: [sourceDocument],
+      targetDocuments: [targetDocument],
+      sourceDocumentRevisions: [sourceRevisionOne, sourceRevisionTwo],
+      targetDocumentRevisions: [targetRevisionOne, targetRevisionTwo],
+      sourceAttachments: [],
+      targetAttachments: [],
+      targetAgents: [],
+      targetProjects: [],
+      targetProjectWorkspaces: [],
+      targetGoals: [{ id: "goal-1" }] as any,
+    });
+
+    expect(plan.counts.documentsToMerge).toBe(1);
+    expect(plan.counts.documentRevisionsToInsert).toBe(1);
+    expect(plan.documentPlans[0]).toMatchObject({
+      action: "merge_existing",
+      latestRevisionId: "revision-branch-2",
+      latestRevisionNumber: 3,
+    });
+    const mergePlan = plan.documentPlans[0] as any;
+    expect(mergePlan.revisionsToInsert).toHaveLength(1);
+    expect(mergePlan.revisionsToInsert[0]).toMatchObject({
+      source: { id: "revision-branch-2" },
+      targetRevisionNumber: 3,
+    });
+  });
+
+  it("imports attachments while clearing missing comment and author references", () => {
+    const sharedIssue = makeIssue({ id: "issue-a", identifier: "PAP-10" });
+    const attachment = makeAttachment({
+      issueId: "issue-a",
+      issueCommentId: "comment-missing",
+      createdByAgentId: "agent-missing",
+    });
+
+    const plan = buildWorktreeMergePlan({
+      companyId: "company-1",
+      companyName: "Paperclip",
+      issuePrefix: "PAP",
+      previewIssueCounterStart: 10,
+      scopes: ["issues"],
+      sourceIssues: [sharedIssue],
+      targetIssues: [sharedIssue],
+      sourceComments: [],
+      targetComments: [],
+      sourceDocuments: [],
+      targetDocuments: [],
+      sourceDocumentRevisions: [],
+      targetDocumentRevisions: [],
+      sourceAttachments: [attachment],
+      targetAttachments: [],
+      targetAgents: [],
+      targetProjects: [],
+      targetProjectWorkspaces: [],
+      targetGoals: [{ id: "goal-1" }] as any,
+    });
+
+    expect(plan.counts.attachmentsToInsert).toBe(1);
+    expect(plan.adjustments.clear_attachment_agent).toBe(1);
+    expect(plan.attachmentPlans[0]).toMatchObject({
+      action: "insert",
+      targetIssueCommentId: null,
+      targetCreatedByAgentId: null,
+    });
+  });
+});

cli/src/__tests__/worktree.test.ts

@@ -0,0 +1,510 @@
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+import { execFileSync } from "node:child_process";
+import { afterEach, describe, expect, it, vi } from "vitest";
+import {
+  copyGitHooksToWorktreeGitDir,
+  copySeededSecretsKey,
+  readSourceAttachmentBody,
+  rebindWorkspaceCwd,
+  resolveSourceConfigPath,
+  resolveGitWorktreeAddArgs,
+  resolveWorktreeMakeTargetPath,
+  worktreeInitCommand,
+  worktreeMakeCommand,
+} from "../commands/worktree.js";
+import {
+  buildWorktreeConfig,
+  buildWorktreeEnvEntries,
+  formatShellExports,
+  generateWorktreeColor,
+  resolveWorktreeSeedPlan,
+  resolveWorktreeLocalPaths,
+  rewriteLocalUrlPort,
+  sanitizeWorktreeInstanceId,
+} from "../commands/worktree-lib.js";
+import type { PaperclipConfig } from "../config/schema.js";
+
+const ORIGINAL_CWD = process.cwd();
+const ORIGINAL_ENV = { ...process.env };
+
+afterEach(() => {
+  process.chdir(ORIGINAL_CWD);
+  for (const key of Object.keys(process.env)) {
+    if (!(key in ORIGINAL_ENV)) delete process.env[key];
+  }
+  for (const [key, value] of Object.entries(ORIGINAL_ENV)) {
+    if (value === undefined) delete process.env[key];
+    else process.env[key] = value;
+  }
+});
+
+function buildSourceConfig(): PaperclipConfig {
+  return {
+    $meta: {
+      version: 1,
+      updatedAt: "2026-03-09T00:00:00.000Z",
+      source: "configure",
+    },
+    database: {
+      mode: "embedded-postgres",
+      embeddedPostgresDataDir: "/tmp/main/db",
+      embeddedPostgresPort: 54329,
+      backup: {
+        enabled: true,
+        intervalMinutes: 60,
+        retentionDays: 30,
+        dir: "/tmp/main/backups",
+      },
+    },
+    logging: {
+      mode: "file",
+      logDir: "/tmp/main/logs",
+    },
+    server: {
+      deploymentMode: "authenticated",
+      exposure: "private",
+      host: "127.0.0.1",
+      port: 3100,
+      allowedHostnames: ["localhost"],
+      serveUi: true,
+    },
+    auth: {
+      baseUrlMode: "explicit",
+      publicBaseUrl: "http://127.0.0.1:3100",
+      disableSignUp: false,
+    },
+    storage: {
+      provider: "local_disk",
+      localDisk: {
+        baseDir: "/tmp/main/storage",
+      },
+      s3: {
+        bucket: "paperclip",
+        region: "us-east-1",
+        prefix: "",
+        forcePathStyle: false,
+      },
+    },
+    secrets: {
+      provider: "local_encrypted",
+      strictMode: false,
+      localEncrypted: {
+        keyFilePath: "/tmp/main/secrets/master.key",
+      },
+    },
+  };
+}
+
+describe("worktree helpers", () => {
+  it("sanitizes instance ids", () => {
+    expect(sanitizeWorktreeInstanceId("feature/worktree-support")).toBe("feature-worktree-support");
+    expect(sanitizeWorktreeInstanceId("  ")).toBe("worktree");
+  });
+
+  it("resolves worktree:make target paths under the user home directory", () => {
+    expect(resolveWorktreeMakeTargetPath("paperclip-pr-432")).toBe(
+      path.resolve(os.homedir(), "paperclip-pr-432"),
+    );
+  });
+
+  it("rejects worktree:make names that are not safe directory/branch names", () => {
+    expect(() => resolveWorktreeMakeTargetPath("paperclip/pr-432")).toThrow(
+      "Worktree name must contain only letters, numbers, dots, underscores, or dashes.",
+    );
+  });
+
+  it("builds git worktree add args for new and existing branches", () => {
+    expect(
+      resolveGitWorktreeAddArgs({
+        branchName: "feature-branch",
+        targetPath: "/tmp/feature-branch",
+        branchExists: false,
+      }),
+    ).toEqual(["worktree", "add", "-b", "feature-branch", "/tmp/feature-branch", "HEAD"]);
+
+    expect(
+      resolveGitWorktreeAddArgs({
+        branchName: "feature-branch",
+        targetPath: "/tmp/feature-branch",
+        branchExists: true,
+      }),
+    ).toEqual(["worktree", "add", "/tmp/feature-branch", "feature-branch"]);
+  });
+
+  it("builds git worktree add args with a start point", () => {
+    expect(
+      resolveGitWorktreeAddArgs({
+        branchName: "my-worktree",
+        targetPath: "/tmp/my-worktree",
+        branchExists: false,
+        startPoint: "public-gh/master",
+      }),
+    ).toEqual(["worktree", "add", "-b", "my-worktree", "/tmp/my-worktree", "public-gh/master"]);
+  });
+
+  it("uses start point even when a local branch with the same name exists", () => {
+    expect(
+      resolveGitWorktreeAddArgs({
+        branchName: "my-worktree",
+        targetPath: "/tmp/my-worktree",
+        branchExists: true,
+        startPoint: "origin/main",
+      }),
+    ).toEqual(["worktree", "add", "-b", "my-worktree", "/tmp/my-worktree", "origin/main"]);
+  });
+
+  it("rewrites loopback auth URLs to the new port only", () => {
+    expect(rewriteLocalUrlPort("http://127.0.0.1:3100", 3110)).toBe("http://127.0.0.1:3110/");
+    expect(rewriteLocalUrlPort("https://paperclip.example", 3110)).toBe("https://paperclip.example");
+  });
+
+  it("builds isolated config and env paths for a worktree", () => {
+    const paths = resolveWorktreeLocalPaths({
+      cwd: "/tmp/paperclip-feature",
+      homeDir: "/tmp/paperclip-worktrees",
+      instanceId: "feature-worktree-support",
+    });
+    const config = buildWorktreeConfig({
+      sourceConfig: buildSourceConfig(),
+      paths,
+      serverPort: 3110,
+      databasePort: 54339,
+      now: new Date("2026-03-09T12:00:00.000Z"),
+    });
+
+    expect(config.database.embeddedPostgresDataDir).toBe(
+      path.resolve("/tmp/paperclip-worktrees", "instances", "feature-worktree-support", "db"),
+    );
+    expect(config.database.embeddedPostgresPort).toBe(54339);
+    expect(config.server.port).toBe(3110);
+    expect(config.auth.publicBaseUrl).toBe("http://127.0.0.1:3110/");
+    expect(config.storage.localDisk.baseDir).toBe(
+      path.resolve("/tmp/paperclip-worktrees", "instances", "feature-worktree-support", "data", "storage"),
+    );
+
+    const env = buildWorktreeEnvEntries(paths, {
+      name: "feature-worktree-support",
+      color: "#3abf7a",
+    });
+    expect(env.PAPERCLIP_HOME).toBe(path.resolve("/tmp/paperclip-worktrees"));
+    expect(env.PAPERCLIP_INSTANCE_ID).toBe("feature-worktree-support");
+    expect(env.PAPERCLIP_IN_WORKTREE).toBe("true");
+    expect(env.PAPERCLIP_WORKTREE_NAME).toBe("feature-worktree-support");
+    expect(env.PAPERCLIP_WORKTREE_COLOR).toBe("#3abf7a");
+    expect(formatShellExports(env)).toContain("export PAPERCLIP_INSTANCE_ID='feature-worktree-support'");
+  });
+
+  it("falls back across storage roots before skipping a missing attachment object", async () => {
+    const missingErr = Object.assign(new Error("missing"), { code: "ENOENT" });
+    const expected = Buffer.from("image-bytes");
+    await expect(
+      readSourceAttachmentBody(
+        [
+          {
+            getObject: vi.fn().mockRejectedValue(missingErr),
+          },
+          {
+            getObject: vi.fn().mockResolvedValue(expected),
+          },
+        ],
+        "company-1",
+        "company-1/issues/issue-1/missing.png",
+      ),
+    ).resolves.toEqual(expected);
+  });
+
+  it("returns null when an attachment object is missing from every lookup storage", async () => {
+    const missingErr = Object.assign(new Error("missing"), { code: "ENOENT" });
+    await expect(
+      readSourceAttachmentBody(
+        [
+          {
+            getObject: vi.fn().mockRejectedValue(missingErr),
+          },
+          {
+            getObject: vi.fn().mockRejectedValue(Object.assign(new Error("missing"), { status: 404 })),
+          },
+        ],
+        "company-1",
+        "company-1/issues/issue-1/missing.png",
+      ),
+    ).resolves.toBeNull();
+  });
+
+  it("generates vivid worktree colors as hex", () => {
+    expect(generateWorktreeColor()).toMatch(/^#[0-9a-f]{6}$/);
+  });
+
+  it("uses minimal seed mode to keep app state but drop heavy runtime history", () => {
+    const minimal = resolveWorktreeSeedPlan("minimal");
+    const full = resolveWorktreeSeedPlan("full");
+
+    expect(minimal.excludedTables).toContain("heartbeat_runs");
+    expect(minimal.excludedTables).toContain("heartbeat_run_events");
+    expect(minimal.excludedTables).toContain("workspace_runtime_services");
+    expect(minimal.excludedTables).toContain("agent_task_sessions");
+    expect(minimal.nullifyColumns.issues).toEqual(["checkout_run_id", "execution_run_id"]);
+
+    expect(full.excludedTables).toEqual([]);
+    expect(full.nullifyColumns).toEqual({});
+  });
+
+  it("copies the source local_encrypted secrets key into the seeded worktree instance", () => {
+    const tempRoot = fs.mkdtempSync(path.join(os.tmpdir(), "paperclip-worktree-secrets-"));
+    const originalInlineMasterKey = process.env.PAPERCLIP_SECRETS_MASTER_KEY;
+    const originalKeyFile = process.env.PAPERCLIP_SECRETS_MASTER_KEY_FILE;
+    try {
+      delete process.env.PAPERCLIP_SECRETS_MASTER_KEY;
+      delete process.env.PAPERCLIP_SECRETS_MASTER_KEY_FILE;
+      const sourceConfigPath = path.join(tempRoot, "source", "config.json");
+      const sourceKeyPath = path.join(tempRoot, "source", "secrets", "master.key");
+      const targetKeyPath = path.join(tempRoot, "target", "secrets", "master.key");
+      fs.mkdirSync(path.dirname(sourceKeyPath), { recursive: true });
+      fs.writeFileSync(sourceKeyPath, "source-master-key", "utf8");
+
+      const sourceConfig = buildSourceConfig();
+      sourceConfig.secrets.localEncrypted.keyFilePath = sourceKeyPath;
+
+      copySeededSecretsKey({
+        sourceConfigPath,
+        sourceConfig,
+        sourceEnvEntries: {},
+        targetKeyFilePath: targetKeyPath,
+      });
+
+      expect(fs.readFileSync(targetKeyPath, "utf8")).toBe("source-master-key");
+    } finally {
+      if (originalInlineMasterKey === undefined) {
+        delete process.env.PAPERCLIP_SECRETS_MASTER_KEY;
+      } else {
+        process.env.PAPERCLIP_SECRETS_MASTER_KEY = originalInlineMasterKey;
+      }
+      if (originalKeyFile === undefined) {
+        delete process.env.PAPERCLIP_SECRETS_MASTER_KEY_FILE;
+      } else {
+        process.env.PAPERCLIP_SECRETS_MASTER_KEY_FILE = originalKeyFile;
+      }
+      fs.rmSync(tempRoot, { recursive: true, force: true });
+    }
+  });
+
+  it("writes the source inline secrets master key into the seeded worktree instance", () => {
+    const tempRoot = fs.mkdtempSync(path.join(os.tmpdir(), "paperclip-worktree-secrets-"));
+    try {
+      const sourceConfigPath = path.join(tempRoot, "source", "config.json");
+      const targetKeyPath = path.join(tempRoot, "target", "secrets", "master.key");
+
+      copySeededSecretsKey({
+        sourceConfigPath,
+        sourceConfig: buildSourceConfig(),
+        sourceEnvEntries: {
+          PAPERCLIP_SECRETS_MASTER_KEY: "inline-source-master-key",
+        },
+        targetKeyFilePath: targetKeyPath,
+      });
+
+      expect(fs.readFileSync(targetKeyPath, "utf8")).toBe("inline-source-master-key");
+    } finally {
+      fs.rmSync(tempRoot, { recursive: true, force: true });
+    }
+  });
+
+  it("persists the current agent jwt secret into the worktree env file", async () => {
+    const tempRoot = fs.mkdtempSync(path.join(os.tmpdir(), "paperclip-worktree-jwt-"));
+    const repoRoot = path.join(tempRoot, "repo");
+    const originalCwd = process.cwd();
+    const originalJwtSecret = process.env.PAPERCLIP_AGENT_JWT_SECRET;
+
+    try {
+      fs.mkdirSync(repoRoot, { recursive: true });
+      process.env.PAPERCLIP_AGENT_JWT_SECRET = "worktree-shared-secret";
+      process.chdir(repoRoot);
+
+      await worktreeInitCommand({
+        seed: false,
+        fromConfig: path.join(tempRoot, "missing", "config.json"),
+        home: path.join(tempRoot, ".paperclip-worktrees"),
+      });
+
+      const envPath = path.join(repoRoot, ".paperclip", ".env");
+      const envContents = fs.readFileSync(envPath, "utf8");
+      expect(envContents).toContain("PAPERCLIP_AGENT_JWT_SECRET=worktree-shared-secret");
+      expect(envContents).toContain("PAPERCLIP_WORKTREE_NAME=repo");
+      expect(envContents).toMatch(/PAPERCLIP_WORKTREE_COLOR=\"#[0-9a-f]{6}\"/);
+    } finally {
+      process.chdir(originalCwd);
+      if (originalJwtSecret === undefined) {
+        delete process.env.PAPERCLIP_AGENT_JWT_SECRET;
+      } else {
+        process.env.PAPERCLIP_AGENT_JWT_SECRET = originalJwtSecret;
+      }
+      fs.rmSync(tempRoot, { recursive: true, force: true });
+    }
+  });
+
+  it("defaults the seed source config to the current repo-local Paperclip config", () => {
+    const tempRoot = fs.mkdtempSync(path.join(os.tmpdir(), "paperclip-worktree-source-config-"));
+    const repoRoot = path.join(tempRoot, "repo");
+    const localConfigPath = path.join(repoRoot, ".paperclip", "config.json");
+    const originalCwd = process.cwd();
+    const originalPaperclipConfig = process.env.PAPERCLIP_CONFIG;
+
+    try {
+      fs.mkdirSync(path.dirname(localConfigPath), { recursive: true });
+      fs.writeFileSync(localConfigPath, JSON.stringify(buildSourceConfig()), "utf8");
+      delete process.env.PAPERCLIP_CONFIG;
+      process.chdir(repoRoot);
+
+      expect(fs.realpathSync(resolveSourceConfigPath({}))).toBe(fs.realpathSync(localConfigPath));
+    } finally {
+      process.chdir(originalCwd);
+      if (originalPaperclipConfig === undefined) {
+        delete process.env.PAPERCLIP_CONFIG;
+      } else {
+        process.env.PAPERCLIP_CONFIG = originalPaperclipConfig;
+      }
+      fs.rmSync(tempRoot, { recursive: true, force: true });
+    }
+  });
+
+  it("preserves the source config path across worktree:make cwd changes", () => {
+    const tempRoot = fs.mkdtempSync(path.join(os.tmpdir(), "paperclip-worktree-source-override-"));
+    const sourceConfigPath = path.join(tempRoot, "source", "config.json");
+    const targetRoot = path.join(tempRoot, "target");
+    const originalCwd = process.cwd();
+    const originalPaperclipConfig = process.env.PAPERCLIP_CONFIG;
+
+    try {
+      fs.mkdirSync(path.dirname(sourceConfigPath), { recursive: true });
+      fs.mkdirSync(targetRoot, { recursive: true });
+      fs.writeFileSync(sourceConfigPath, JSON.stringify(buildSourceConfig()), "utf8");
+      delete process.env.PAPERCLIP_CONFIG;
+      process.chdir(targetRoot);
+
+      expect(resolveSourceConfigPath({ sourceConfigPathOverride: sourceConfigPath })).toBe(
+        path.resolve(sourceConfigPath),
+      );
+    } finally {
+      process.chdir(originalCwd);
+      if (originalPaperclipConfig === undefined) {
+        delete process.env.PAPERCLIP_CONFIG;
+      } else {
+        process.env.PAPERCLIP_CONFIG = originalPaperclipConfig;
+      }
+      fs.rmSync(tempRoot, { recursive: true, force: true });
+    }
+  });
+
+  it("rebinds same-repo workspace paths onto the current worktree root", () => {
+    expect(
+      rebindWorkspaceCwd({
+        sourceRepoRoot: "/Users/example/paperclip",
+        targetRepoRoot: "/Users/example/paperclip-pr-432",
+        workspaceCwd: "/Users/example/paperclip",
+      }),
+    ).toBe("/Users/example/paperclip-pr-432");
+
+    expect(
+      rebindWorkspaceCwd({
+        sourceRepoRoot: "/Users/example/paperclip",
+        targetRepoRoot: "/Users/example/paperclip-pr-432",
+        workspaceCwd: "/Users/example/paperclip/packages/db",
+      }),
+    ).toBe("/Users/example/paperclip-pr-432/packages/db");
+  });
+
+  it("does not rebind paths outside the source repo root", () => {
+    expect(
+      rebindWorkspaceCwd({
+        sourceRepoRoot: "/Users/example/paperclip",
+        targetRepoRoot: "/Users/example/paperclip-pr-432",
+        workspaceCwd: "/Users/example/other-project",
+      }),
+    ).toBeNull();
+  });
+
+  it("copies shared git hooks into a linked worktree git dir", () => {
+    const tempRoot = fs.mkdtempSync(path.join(os.tmpdir(), "paperclip-worktree-hooks-"));
+    const repoRoot = path.join(tempRoot, "repo");
+    const worktreePath = path.join(tempRoot, "repo-feature");
+
+    try {
+      fs.mkdirSync(repoRoot, { recursive: true });
+      execFileSync("git", ["init"], { cwd: repoRoot, stdio: "ignore" });
+      execFileSync("git", ["config", "user.email", "test@example.com"], { cwd: repoRoot, stdio: "ignore" });
+      execFileSync("git", ["config", "user.name", "Test User"], { cwd: repoRoot, stdio: "ignore" });
+      fs.writeFileSync(path.join(repoRoot, "README.md"), "# temp\n", "utf8");
+      execFileSync("git", ["add", "README.md"], { cwd: repoRoot, stdio: "ignore" });
+      execFileSync("git", ["commit", "-m", "Initial commit"], { cwd: repoRoot, stdio: "ignore" });
+
+      const sourceHooksDir = path.join(repoRoot, ".git", "hooks");
+      const sourceHookPath = path.join(sourceHooksDir, "pre-commit");
+      const sourceTokensPath = path.join(sourceHooksDir, "forbidden-tokens.txt");
+      fs.writeFileSync(sourceHookPath, "#!/usr/bin/env bash\nexit 0\n", { encoding: "utf8", mode: 0o755 });
+      fs.chmodSync(sourceHookPath, 0o755);
+      fs.writeFileSync(sourceTokensPath, "secret-token\n", "utf8");
+
+      execFileSync("git", ["worktree", "add", "--detach", worktreePath], { cwd: repoRoot, stdio: "ignore" });
+
+      const copied = copyGitHooksToWorktreeGitDir(worktreePath);
+      const worktreeGitDir = execFileSync("git", ["rev-parse", "--git-dir"], {
+        cwd: worktreePath,
+        encoding: "utf8",
+        stdio: ["ignore", "pipe", "ignore"],
+      }).trim();
+      const resolvedSourceHooksDir = fs.realpathSync(sourceHooksDir);
+      const resolvedTargetHooksDir = fs.realpathSync(path.resolve(worktreePath, worktreeGitDir, "hooks"));
+      const targetHookPath = path.join(resolvedTargetHooksDir, "pre-commit");
+      const targetTokensPath = path.join(resolvedTargetHooksDir, "forbidden-tokens.txt");
+
+      expect(copied).toMatchObject({
+        sourceHooksPath: resolvedSourceHooksDir,
+        targetHooksPath: resolvedTargetHooksDir,
+        copied: true,
+      });
+      expect(fs.readFileSync(targetHookPath, "utf8")).toBe("#!/usr/bin/env bash\nexit 0\n");
+      expect(fs.statSync(targetHookPath).mode & 0o111).not.toBe(0);
+      expect(fs.readFileSync(targetTokensPath, "utf8")).toBe("secret-token\n");
+    } finally {
+      execFileSync("git", ["worktree", "remove", "--force", worktreePath], { cwd: repoRoot, stdio: "ignore" });
+      fs.rmSync(tempRoot, { recursive: true, force: true });
+    }
+  });
+
+  it("creates and initializes a worktree from the top-level worktree:make command", async () => {
+    const tempRoot = fs.mkdtempSync(path.join(os.tmpdir(), "paperclip-worktree-make-"));
+    const repoRoot = path.join(tempRoot, "repo");
+    const fakeHome = path.join(tempRoot, "home");
+    const worktreePath = path.join(fakeHome, "paperclip-make-test");
+    const originalCwd = process.cwd();
+    const homedirSpy = vi.spyOn(os, "homedir").mockReturnValue(fakeHome);
+
+    try {
+      fs.mkdirSync(repoRoot, { recursive: true });
+      fs.mkdirSync(fakeHome, { recursive: true });
+      execFileSync("git", ["init"], { cwd: repoRoot, stdio: "ignore" });
+      execFileSync("git", ["config", "user.email", "test@example.com"], { cwd: repoRoot, stdio: "ignore" });
+      execFileSync("git", ["config", "user.name", "Test User"], { cwd: repoRoot, stdio: "ignore" });
+      fs.writeFileSync(path.join(repoRoot, "README.md"), "# temp\n", "utf8");
+      execFileSync("git", ["add", "README.md"], { cwd: repoRoot, stdio: "ignore" });
+      execFileSync("git", ["commit", "-m", "Initial commit"], { cwd: repoRoot, stdio: "ignore" });
+
+      process.chdir(repoRoot);
+
+      await worktreeMakeCommand("paperclip-make-test", {
+        seed: false,
+        home: path.join(tempRoot, ".paperclip-worktrees"),
+      });
+
+      expect(fs.existsSync(path.join(worktreePath, ".git"))).toBe(true);
+      expect(fs.existsSync(path.join(worktreePath, ".paperclip", "config.json"))).toBe(true);
+      expect(fs.existsSync(path.join(worktreePath, ".paperclip", ".env"))).toBe(true);
+    } finally {
+      process.chdir(originalCwd);
+      homedirSpy.mockRestore();
+      fs.rmSync(tempRoot, { recursive: true, force: true });
+    }
+  }, 20_000);
+});

cli/src/adapters/registry.ts

@@ -1,7 +1,11 @@
 import type { CLIAdapterModule } from "@paperclipai/adapter-utils";
 import { printClaudeStreamEvent } from "@paperclipai/adapter-claude-local/cli";
 import { printCodexStreamEvent } from "@paperclipai/adapter-codex-local/cli";
-import { printOpenClawStreamEvent } from "@paperclipai/adapter-openclaw/cli";
+import { printCursorStreamEvent } from "@paperclipai/adapter-cursor-local/cli";
+import { printGeminiStreamEvent } from "@paperclipai/adapter-gemini-local/cli";
+import { printOpenCodeStreamEvent } from "@paperclipai/adapter-opencode-local/cli";
+import { printPiStreamEvent } from "@paperclipai/adapter-pi-local/cli";
+import { printOpenClawGatewayStreamEvent } from "@paperclipai/adapter-openclaw-gateway/cli";
 import { processCLIAdapter } from "./process/index.js";
 import { httpCLIAdapter } from "./http/index.js";
 
@@ -15,13 +19,43 @@ const codexLocalCLIAdapter: CLIAdapterModule = {
   formatStdoutEvent: printCodexStreamEvent,
 };
 
-const openclawCLIAdapter: CLIAdapterModule = {
-  type: "openclaw",
-  formatStdoutEvent: printOpenClawStreamEvent,
+const openCodeLocalCLIAdapter: CLIAdapterModule = {
+  type: "opencode_local",
+  formatStdoutEvent: printOpenCodeStreamEvent,
+};
+
+const piLocalCLIAdapter: CLIAdapterModule = {
+  type: "pi_local",
+  formatStdoutEvent: printPiStreamEvent,
+};
+
+const cursorLocalCLIAdapter: CLIAdapterModule = {
+  type: "cursor",
+  formatStdoutEvent: printCursorStreamEvent,
+};
+
+const geminiLocalCLIAdapter: CLIAdapterModule = {
+  type: "gemini_local",
+  formatStdoutEvent: printGeminiStreamEvent,
+};
+
+const openclawGatewayCLIAdapter: CLIAdapterModule = {
+  type: "openclaw_gateway",
+  formatStdoutEvent: printOpenClawGatewayStreamEvent,
 };
 
 const adaptersByType = new Map<string, CLIAdapterModule>(
-  [claudeLocalCLIAdapter, codexLocalCLIAdapter, openclawCLIAdapter, processCLIAdapter, httpCLIAdapter].map((a) => [a.type, a]),
+  [
+    claudeLocalCLIAdapter,
+    codexLocalCLIAdapter,
+    openCodeLocalCLIAdapter,
+    piLocalCLIAdapter,
+    cursorLocalCLIAdapter,
+    geminiLocalCLIAdapter,
+    openclawGatewayCLIAdapter,
+    processCLIAdapter,
+    httpCLIAdapter,
+  ].map((a) => [a.type, a]),
 );
 
 export function getCLIAdapter(type: string): CLIAdapterModule {

cli/src/client/http.ts

@@ -104,8 +104,10 @@ export class PaperclipApiClient {
 
 function buildUrl(apiBase: string, path: string): string {
   const normalizedPath = path.startsWith("/") ? path : `/${path}`;
+  const [pathname, query] = normalizedPath.split("?");
   const url = new URL(apiBase);
-  url.pathname = `${url.pathname.replace(/\/+$/, "")}${normalizedPath}`;
+  url.pathname = `${url.pathname.replace(/\/+$/, "")}${pathname}`;
+  if (query) url.search = query;
   return url.toString();
 }
 

cli/src/commands/allowed-hostname.ts

@@ -26,6 +26,9 @@ export async function addAllowedHostname(host: string, opts: { config?: string }
     p.log.info(`Hostname ${pc.cyan(normalized)} is already allowed.`);
   } else {
     p.log.success(`Added allowed hostname: ${pc.cyan(normalized)}`);
+    p.log.message(
+      pc.dim("Restart the Paperclip server for this change to take effect."),
+    );
   }
 
   if (!(config.server.deploymentMode === "authenticated" && config.server.exposure === "private")) {

cli/src/commands/auth-bootstrap-ceo.ts

@@ -3,6 +3,7 @@ import * as p from "@clack/prompts";
 import pc from "picocolors";
 import { and, eq, gt, isNull } from "drizzle-orm";
 import { createDb, instanceUserRoles, invites } from "@paperclipai/db";
+import { loadPaperclipEnvFile } from "../config/env.js";
 import { readConfig, resolveConfigPath } from "../config/store.js";
 
 function hashToken(token: string) {
@@ -13,7 +14,8 @@ function createInviteToken() {
   return `pcp_bootstrap_${randomBytes(24).toString("hex")}`;
 }
 
-function resolveDbUrl(configPath?: string) {
+function resolveDbUrl(configPath?: string, explicitDbUrl?: string) {
+  if (explicitDbUrl) return explicitDbUrl;
   const config = readConfig(configPath);
   if (process.env.DATABASE_URL) return process.env.DATABASE_URL;
   if (config?.database.mode === "postgres" && config.database.connectionString) {
@@ -28,6 +30,12 @@ function resolveDbUrl(configPath?: string) {
 
 function resolveBaseUrl(configPath?: string, explicitBaseUrl?: string) {
   if (explicitBaseUrl) return explicitBaseUrl.replace(/\/+$/, "");
+  const fromEnv =
+    process.env.PAPERCLIP_PUBLIC_URL ??
+    process.env.PAPERCLIP_AUTH_PUBLIC_BASE_URL ??
+    process.env.BETTER_AUTH_URL ??
+    process.env.BETTER_AUTH_BASE_URL;
+  if (fromEnv?.trim()) return fromEnv.trim().replace(/\/+$/, "");
   const config = readConfig(configPath);
   if (config?.auth.baseUrlMode === "explicit" && config.auth.publicBaseUrl) {
     return config.auth.publicBaseUrl.replace(/\/+$/, "");
@@ -43,8 +51,10 @@ export async function bootstrapCeoInvite(opts: {
   force?: boolean;
   expiresHours?: number;
   baseUrl?: string;
+  dbUrl?: string;
 }) {
   const configPath = resolveConfigPath(opts.config);
+  loadPaperclipEnvFile(configPath);
   const config = readConfig(configPath);
   if (!config) {
     p.log.error(`No config found at ${configPath}. Run ${pc.cyan("paperclip onboard")} first.`);
@@ -56,7 +66,7 @@ export async function bootstrapCeoInvite(opts: {
     return;
   }
 
-  const dbUrl = resolveDbUrl(configPath);
+  const dbUrl = resolveDbUrl(configPath, opts.dbUrl);
   if (!dbUrl) {
     p.log.error(
       "Could not resolve database connection for bootstrap.",
@@ -65,6 +75,11 @@ export async function bootstrapCeoInvite(opts: {
   }
 
   const db = createDb(dbUrl);
+  const closableDb = db as typeof db & {
+    $client?: {
+      end?: (options?: { timeout?: number }) => Promise<void>;
+    };
+  };
   try {
     const existingAdminCount = await db
       .select()
@@ -112,5 +127,7 @@ export async function bootstrapCeoInvite(opts: {
   } catch (err) {
     p.log.error(`Could not create bootstrap invite: ${err instanceof Error ? err.message : String(err)}`);
     p.log.info("If using embedded-postgres, start the Paperclip server and run this command again.");
+  } finally {
+    await closableDb.$client?.end?.({ timeout: 5 }).catch(() => undefined);
   }
 }

cli/src/commands/client/agent.ts

@@ -1,5 +1,13 @@
 import { Command } from "commander";
 import type { Agent } from "@paperclipai/shared";
+import {
+  removeMaintainerOnlySkillSymlinks,
+  resolvePaperclipSkillsDir,
+} from "@paperclipai/adapter-utils/server-utils";
+import fs from "node:fs/promises";
+import os from "node:os";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
 import {
   addCommonClientOptions,
   formatInlineRecord,
@@ -13,6 +21,141 @@ interface AgentListOptions extends BaseClientOptions {
   companyId?: string;
 }
 
+interface AgentLocalCliOptions extends BaseClientOptions {
+  companyId?: string;
+  keyName?: string;
+  installSkills?: boolean;
+}
+
+interface CreatedAgentKey {
+  id: string;
+  name: string;
+  token: string;
+  createdAt: string;
+}
+
+interface SkillsInstallSummary {
+  tool: "codex" | "claude";
+  target: string;
+  linked: string[];
+  removed: string[];
+  skipped: string[];
+  failed: Array<{ name: string; error: string }>;
+}
+
+const __moduleDir = path.dirname(fileURLToPath(import.meta.url));
+
+function codexSkillsHome(): string {
+  const fromEnv = process.env.CODEX_HOME?.trim();
+  const base = fromEnv && fromEnv.length > 0 ? fromEnv : path.join(os.homedir(), ".codex");
+  return path.join(base, "skills");
+}
+
+function claudeSkillsHome(): string {
+  const fromEnv = process.env.CLAUDE_HOME?.trim();
+  const base = fromEnv && fromEnv.length > 0 ? fromEnv : path.join(os.homedir(), ".claude");
+  return path.join(base, "skills");
+}
+
+async function installSkillsForTarget(
+  sourceSkillsDir: string,
+  targetSkillsDir: string,
+  tool: "codex" | "claude",
+): Promise<SkillsInstallSummary> {
+  const summary: SkillsInstallSummary = {
+    tool,
+    target: targetSkillsDir,
+    linked: [],
+    removed: [],
+    skipped: [],
+    failed: [],
+  };
+
+  await fs.mkdir(targetSkillsDir, { recursive: true });
+  const entries = await fs.readdir(sourceSkillsDir, { withFileTypes: true });
+  summary.removed = await removeMaintainerOnlySkillSymlinks(
+    targetSkillsDir,
+    entries.filter((entry) => entry.isDirectory()).map((entry) => entry.name),
+  );
+  for (const entry of entries) {
+    if (!entry.isDirectory()) continue;
+    const source = path.join(sourceSkillsDir, entry.name);
+    const target = path.join(targetSkillsDir, entry.name);
+    const existing = await fs.lstat(target).catch(() => null);
+    if (existing) {
+      if (existing.isSymbolicLink()) {
+        let linkedPath: string | null = null;
+        try {
+          linkedPath = await fs.readlink(target);
+        } catch (err) {
+          await fs.unlink(target);
+          try {
+            await fs.symlink(source, target);
+            summary.linked.push(entry.name);
+            continue;
+          } catch (linkErr) {
+            summary.failed.push({
+              name: entry.name,
+              error:
+                err instanceof Error && linkErr instanceof Error
+                  ? `${err.message}; then ${linkErr.message}`
+                  : err instanceof Error
+                    ? err.message
+                    : `Failed to recover broken symlink: ${String(err)}`,
+            });
+            continue;
+          }
+        }
+
+        const resolvedLinkedPath = path.isAbsolute(linkedPath)
+          ? linkedPath
+          : path.resolve(path.dirname(target), linkedPath);
+        const linkedTargetExists = await fs
+          .stat(resolvedLinkedPath)
+          .then(() => true)
+          .catch(() => false);
+
+        if (!linkedTargetExists) {
+          await fs.unlink(target);
+        } else {
+          summary.skipped.push(entry.name);
+          continue;
+        }
+      } else {
+        summary.skipped.push(entry.name);
+        continue;
+      }
+    }
+
+    try {
+      await fs.symlink(source, target);
+      summary.linked.push(entry.name);
+    } catch (err) {
+      summary.failed.push({
+        name: entry.name,
+        error: err instanceof Error ? err.message : String(err),
+      });
+    }
+  }
+
+  return summary;
+}
+
+function buildAgentEnvExports(input: {
+  apiBase: string;
+  companyId: string;
+  agentId: string;
+  apiKey: string;
+}): string {
+  const escaped = (value: string) => value.replace(/'/g, "'\"'\"'");
+  return [
+    `export PAPERCLIP_API_URL='${escaped(input.apiBase)}'`,
+    `export PAPERCLIP_COMPANY_ID='${escaped(input.companyId)}'`,
+    `export PAPERCLIP_AGENT_ID='${escaped(input.agentId)}'`,
+    `export PAPERCLIP_API_KEY='${escaped(input.apiKey)}'`,
+  ].join("\n");
+}
+
 export function registerAgentCommands(program: Command): void {
   const agent = program.command("agent").description("Agent operations");
 
@@ -71,4 +214,102 @@ export function registerAgentCommands(program: Command): void {
         }
       }),
   );
+
+  addCommonClientOptions(
+    agent
+      .command("local-cli")
+      .description(
+        "Create an agent API key, install local Paperclip skills for Codex/Claude, and print shell exports",
+      )
+      .argument("<agentRef>", "Agent ID or shortname/url-key")
+      .requiredOption("-C, --company-id <id>", "Company ID")
+      .option("--key-name <name>", "API key label", "local-cli")
+      .option(
+        "--no-install-skills",
+        "Skip installing Paperclip skills into ~/.codex/skills and ~/.claude/skills",
+      )
+      .action(async (agentRef: string, opts: AgentLocalCliOptions) => {
+        try {
+          const ctx = resolveCommandContext(opts, { requireCompany: true });
+          const query = new URLSearchParams({ companyId: ctx.companyId ?? "" });
+          const agentRow = await ctx.api.get<Agent>(
+            `/api/agents/${encodeURIComponent(agentRef)}?${query.toString()}`,
+          );
+          if (!agentRow) {
+            throw new Error(`Agent not found: ${agentRef}`);
+          }
+
+          const now = new Date().toISOString().replaceAll(":", "-");
+          const keyName = opts.keyName?.trim() ? opts.keyName.trim() : `local-cli-${now}`;
+          const key = await ctx.api.post<CreatedAgentKey>(`/api/agents/${agentRow.id}/keys`, { name: keyName });
+          if (!key) {
+            throw new Error("Failed to create API key");
+          }
+
+          const installSummaries: SkillsInstallSummary[] = [];
+          if (opts.installSkills !== false) {
+            const skillsDir = await resolvePaperclipSkillsDir(__moduleDir, [path.resolve(process.cwd(), "skills")]);
+            if (!skillsDir) {
+              throw new Error(
+                "Could not locate local Paperclip skills directory. Expected ./skills in the repo checkout.",
+              );
+            }
+
+            installSummaries.push(
+              await installSkillsForTarget(skillsDir, codexSkillsHome(), "codex"),
+              await installSkillsForTarget(skillsDir, claudeSkillsHome(), "claude"),
+            );
+          }
+
+          const exportsText = buildAgentEnvExports({
+            apiBase: ctx.api.apiBase,
+            companyId: agentRow.companyId,
+            agentId: agentRow.id,
+            apiKey: key.token,
+          });
+
+          if (ctx.json) {
+            printOutput(
+              {
+                agent: {
+                  id: agentRow.id,
+                  name: agentRow.name,
+                  urlKey: agentRow.urlKey,
+                  companyId: agentRow.companyId,
+                },
+                key: {
+                  id: key.id,
+                  name: key.name,
+                  createdAt: key.createdAt,
+                  token: key.token,
+                },
+                skills: installSummaries,
+                exports: exportsText,
+              },
+              { json: true },
+            );
+            return;
+          }
+
+          console.log(`Agent: ${agentRow.name} (${agentRow.id})`);
+          console.log(`API key created: ${key.name} (${key.id})`);
+          if (installSummaries.length > 0) {
+            for (const summary of installSummaries) {
+              console.log(
+                `${summary.tool}: linked=${summary.linked.length} removed=${summary.removed.length} skipped=${summary.skipped.length} failed=${summary.failed.length} target=${summary.target}`,
+              );
+              for (const failed of summary.failed) {
+                console.log(`  failed ${failed.name}: ${failed.error}`);
+              }
+            }
+          }
+          console.log("");
+          console.log("# Run this in your shell before launching codex/claude:");
+          console.log(exportsText);
+        } catch (err) {
+          handleCommandError(err);
+        }
+      }),
+    { includeCompany: false },
+  );
 }

cli/src/commands/client/company.ts

@@ -1,11 +1,12 @@
 import { Command } from "commander";
-import { mkdir, readFile, stat, writeFile } from "node:fs/promises";
+import { mkdir, readdir, readFile, stat, writeFile } from "node:fs/promises";
 import path from "node:path";
+import * as p from "@clack/prompts";
 import type {
   Company,
+  CompanyPortabilityFileEntry,
   CompanyPortabilityExportResult,
   CompanyPortabilityInclude,
-  CompanyPortabilityManifest,
   CompanyPortabilityPreviewResult,
   CompanyPortabilityImportResult,
 } from "@paperclipai/shared";
@@ -33,6 +34,11 @@ interface CompanyDeleteOptions extends BaseClientOptions {
 interface CompanyExportOptions extends BaseClientOptions {
   out?: string;
   include?: string;
+  skills?: string;
+  projects?: string;
+  issues?: string;
+  projectIssues?: string;
+  expandReferencedSkills?: boolean;
 }
 
 interface CompanyImportOptions extends BaseClientOptions {
@@ -46,6 +52,30 @@ interface CompanyImportOptions extends BaseClientOptions {
   dryRun?: boolean;
 }
 
+const binaryContentTypeByExtension: Record<string, string> = {
+  ".gif": "image/gif",
+  ".jpeg": "image/jpeg",
+  ".jpg": "image/jpeg",
+  ".png": "image/png",
+  ".svg": "image/svg+xml",
+  ".webp": "image/webp",
+};
+
+function readPortableFileEntry(filePath: string, contents: Buffer): CompanyPortabilityFileEntry {
+  const contentType = binaryContentTypeByExtension[path.extname(filePath).toLowerCase()];
+  if (!contentType) return contents.toString("utf8");
+  return {
+    encoding: "base64",
+    data: contents.toString("base64"),
+    contentType,
+  };
+}
+
+function portableFileEntryToWriteValue(entry: CompanyPortabilityFileEntry): string | Uint8Array {
+  if (typeof entry === "string") return entry;
+  return Buffer.from(entry.data, "base64");
+}
+
 function isUuidLike(value: string): boolean {
   return /^[0-9a-f]{8}-[0-9a-f]{4}-[1-5][0-9a-f]{3}-[89ab][0-9a-f]{3}-[0-9a-f]{12}$/i.test(value);
 }
@@ -55,14 +85,17 @@ function normalizeSelector(input: string): string {
 }
 
 function parseInclude(input: string | undefined): CompanyPortabilityInclude {
-  if (!input || !input.trim()) return { company: true, agents: true };
+  if (!input || !input.trim()) return { company: true, agents: true, projects: false, issues: false, skills: false };
   const values = input.split(",").map((part) => part.trim().toLowerCase()).filter(Boolean);
   const include = {
     company: values.includes("company"),
     agents: values.includes("agents"),
+    projects: values.includes("projects"),
+    issues: values.includes("issues") || values.includes("tasks"),
+    skills: values.includes("skills"),
   };
-  if (!include.company && !include.agents) {
-    throw new Error("Invalid --include value. Use one or both of: company,agents");
+  if (!include.company && !include.agents && !include.projects && !include.issues && !include.skills) {
+    throw new Error("Invalid --include value. Use one or more of: company,agents,projects,issues,tasks,skills");
   }
   return include;
 }
@@ -76,50 +109,95 @@ function parseAgents(input: string | undefined): "all" | string[] {
   return Array.from(new Set(values));
 }
 
-function isHttpUrl(input: string): boolean {
+function parseCsvValues(input: string | undefined): string[] {
+  if (!input || !input.trim()) return [];
+  return Array.from(new Set(input.split(",").map((part) => part.trim()).filter(Boolean)));
+}
+
+export function isHttpUrl(input: string): boolean {
   return /^https?:\/\//i.test(input.trim());
 }
 
-function isGithubUrl(input: string): boolean {
+export function isGithubUrl(input: string): boolean {
   return /^https?:\/\/github\.com\//i.test(input.trim());
 }
 
+async function collectPackageFiles(
+  root: string,
+  current: string,
+  files: Record<string, CompanyPortabilityFileEntry>,
+): Promise<void> {
+  const entries = await readdir(current, { withFileTypes: true });
+  for (const entry of entries) {
+    if (entry.name.startsWith(".git")) continue;
+    const absolutePath = path.join(current, entry.name);
+    if (entry.isDirectory()) {
+      await collectPackageFiles(root, absolutePath, files);
+      continue;
+    }
+    if (!entry.isFile()) continue;
+    const isMarkdown = entry.name.endsWith(".md");
+    const isPaperclipYaml = entry.name === ".paperclip.yaml" || entry.name === ".paperclip.yml";
+    const contentType = binaryContentTypeByExtension[path.extname(entry.name).toLowerCase()];
+    if (!isMarkdown && !isPaperclipYaml && !contentType) continue;
+    const relativePath = path.relative(root, absolutePath).replace(/\\/g, "/");
+    files[relativePath] = readPortableFileEntry(relativePath, await readFile(absolutePath));
+  }
+}
+
 async function resolveInlineSourceFromPath(inputPath: string): Promise<{
-  manifest: CompanyPortabilityManifest;
-  files: Record<string, string>;
+  rootPath: string;
+  files: Record<string, CompanyPortabilityFileEntry>;
 }> {
   const resolved = path.resolve(inputPath);
   const resolvedStat = await stat(resolved);
-  const manifestPath = resolvedStat.isDirectory()
-    ? path.join(resolved, "paperclip.manifest.json")
-    : resolved;
-  const manifestBaseDir = path.dirname(manifestPath);
-  const manifestRaw = await readFile(manifestPath, "utf8");
-  const manifest = JSON.parse(manifestRaw) as CompanyPortabilityManifest;
-  const files: Record<string, string> = {};
-
-  if (manifest.company?.path) {
-    const companyPath = manifest.company.path.replace(/\\/g, "/");
-    files[companyPath] = await readFile(path.join(manifestBaseDir, companyPath), "utf8");
-  }
-  for (const agent of manifest.agents ?? []) {
-    const agentPath = agent.path.replace(/\\/g, "/");
-    files[agentPath] = await readFile(path.join(manifestBaseDir, agentPath), "utf8");
-  }
-
-  return { manifest, files };
+  const rootDir = resolvedStat.isDirectory() ? resolved : path.dirname(resolved);
+  const files: Record<string, CompanyPortabilityFileEntry> = {};
+  await collectPackageFiles(rootDir, rootDir, files);
+  return {
+    rootPath: path.basename(rootDir),
+    files,
+  };
 }
 
 async function writeExportToFolder(outDir: string, exported: CompanyPortabilityExportResult): Promise<void> {
   const root = path.resolve(outDir);
   await mkdir(root, { recursive: true });
-  const manifestPath = path.join(root, "paperclip.manifest.json");
-  await writeFile(manifestPath, JSON.stringify(exported.manifest, null, 2), "utf8");
   for (const [relativePath, content] of Object.entries(exported.files)) {
     const normalized = relativePath.replace(/\\/g, "/");
     const filePath = path.join(root, normalized);
     await mkdir(path.dirname(filePath), { recursive: true });
-    await writeFile(filePath, content, "utf8");
+    const writeValue = portableFileEntryToWriteValue(content);
+    if (typeof writeValue === "string") {
+      await writeFile(filePath, writeValue, "utf8");
+    } else {
+      await writeFile(filePath, writeValue);
+    }
+  }
+}
+
+async function confirmOverwriteExportDirectory(outDir: string): Promise<void> {
+  const root = path.resolve(outDir);
+  const stats = await stat(root).catch(() => null);
+  if (!stats) return;
+  if (!stats.isDirectory()) {
+    throw new Error(`Export output path ${root} exists and is not a directory.`);
+  }
+
+  const entries = await readdir(root);
+  if (entries.length === 0) return;
+
+  if (!process.stdin.isTTY || !process.stdout.isTTY) {
+    throw new Error(`Export output directory ${root} already contains files. Re-run interactively or choose an empty directory.`);
+  }
+
+  const confirmed = await p.confirm({
+    message: `Overwrite existing files in ${root}?`,
+    initialValue: false,
+  });
+
+  if (p.isCancel(confirmed) || !confirmed) {
+    throw new Error("Export cancelled.");
   }
 }
 
@@ -257,27 +335,42 @@ export function registerCompanyCommands(program: Command): void {
   addCommonClientOptions(
     company
       .command("export")
-      .description("Export a company into portable manifest + markdown files")
+      .description("Export a company into a portable markdown package")
       .argument("<companyId>", "Company ID")
       .requiredOption("--out <path>", "Output directory")
-      .option("--include <values>", "Comma-separated include set: company,agents", "company,agents")
+      .option("--include <values>", "Comma-separated include set: company,agents,projects,issues,tasks,skills", "company,agents")
+      .option("--skills <values>", "Comma-separated skill slugs/keys to export")
+      .option("--projects <values>", "Comma-separated project shortnames/ids to export")
+      .option("--issues <values>", "Comma-separated issue identifiers/ids to export")
+      .option("--project-issues <values>", "Comma-separated project shortnames/ids whose issues should be exported")
+      .option("--expand-referenced-skills", "Vendor skill contents instead of exporting upstream references", false)
       .action(async (companyId: string, opts: CompanyExportOptions) => {
         try {
           const ctx = resolveCommandContext(opts);
           const include = parseInclude(opts.include);
           const exported = await ctx.api.post<CompanyPortabilityExportResult>(
             `/api/companies/${companyId}/export`,
-            { include },
+            {
+              include,
+              skills: parseCsvValues(opts.skills),
+              projects: parseCsvValues(opts.projects),
+              issues: parseCsvValues(opts.issues),
+              projectIssues: parseCsvValues(opts.projectIssues),
+              expandReferencedSkills: Boolean(opts.expandReferencedSkills),
+            },
           );
           if (!exported) {
             throw new Error("Export request returned no data");
           }
+          await confirmOverwriteExportDirectory(opts.out!);
           await writeExportToFolder(opts.out!, exported);
           printOutput(
             {
               ok: true,
               out: path.resolve(opts.out!),
-              filesWritten: Object.keys(exported.files).length + 1,
+              rootPath: exported.rootPath,
+              filesWritten: Object.keys(exported.files).length,
+              paperclipExtensionPath: exported.paperclipExtensionPath,
               warningCount: exported.warnings.length,
             },
             { json: ctx.json },
@@ -296,9 +389,9 @@ export function registerCompanyCommands(program: Command): void {
   addCommonClientOptions(
     company
       .command("import")
-      .description("Import a portable company package from local path, URL, or GitHub")
+      .description("Import a portable markdown company package from local path, URL, or GitHub")
       .requiredOption("--from <pathOrUrl>", "Source path or URL")
-      .option("--include <values>", "Comma-separated include set: company,agents", "company,agents")
+      .option("--include <values>", "Comma-separated include set: company,agents,projects,issues,tasks,skills", "company,agents")
       .option("--target <mode>", "Target mode: new | existing")
       .option("-C, --company-id <id>", "Existing target company ID")
       .option("--new-company-name <name>", "Name override for --target new")
@@ -343,19 +436,22 @@ export function registerCompanyCommands(program: Command): void {
           }
 
           let sourcePayload:
-            | { type: "inline"; manifest: CompanyPortabilityManifest; files: Record<string, string> }
-            | { type: "url"; url: string }
+            | { type: "inline"; rootPath?: string | null; files: Record<string, CompanyPortabilityFileEntry> }
             | { type: "github"; url: string };
 
           if (isHttpUrl(from)) {
-            sourcePayload = isGithubUrl(from)
-              ? { type: "github", url: from }
-              : { type: "url", url: from };
+            if (!isGithubUrl(from)) {
+              throw new Error(
+                "Only GitHub URLs and local paths are supported for import. " +
+                "Generic HTTP URLs are not supported. Use a GitHub URL (https://github.com/...) or a local directory path.",
+              );
+            }
+            sourcePayload = { type: "github", url: from };
           } else {
             const inline = await resolveInlineSourceFromPath(from);
             sourcePayload = {
               type: "inline",
-              manifest: inline.manifest,
+              rootPath: inline.rootPath,
               files: inline.files,
             };
           }

cli/src/commands/client/plugin.ts

@@ -0,0 +1,374 @@
+import path from "node:path";
+import { Command } from "commander";
+import pc from "picocolors";
+import {
+  addCommonClientOptions,
+  handleCommandError,
+  printOutput,
+  resolveCommandContext,
+  type BaseClientOptions,
+} from "./common.js";
+
+// ---------------------------------------------------------------------------
+// Types mirroring server-side shapes
+// ---------------------------------------------------------------------------
+
+interface PluginRecord {
+  id: string;
+  pluginKey: string;
+  packageName: string;
+  version: string;
+  status: string;
+  displayName?: string;
+  lastError?: string | null;
+  installedAt: string;
+  updatedAt: string;
+}
+
+
+// ---------------------------------------------------------------------------
+// Option types
+// ---------------------------------------------------------------------------
+
+interface PluginListOptions extends BaseClientOptions {
+  status?: string;
+}
+
+interface PluginInstallOptions extends BaseClientOptions {
+  local?: boolean;
+  version?: string;
+}
+
+interface PluginUninstallOptions extends BaseClientOptions {
+  force?: boolean;
+}
+
+// ---------------------------------------------------------------------------
+// Helpers
+// ---------------------------------------------------------------------------
+
+/**
+ * Resolve a local path argument to an absolute path so the server can find the
+ * plugin on disk regardless of where the user ran the CLI.
+ */
+function resolvePackageArg(packageArg: string, isLocal: boolean): string {
+  if (!isLocal) return packageArg;
+  // Already absolute
+  if (path.isAbsolute(packageArg)) return packageArg;
+  // Expand leading ~ to home directory
+  if (packageArg.startsWith("~")) {
+    const home = process.env.HOME ?? process.env.USERPROFILE ?? "";
+    return path.resolve(home, packageArg.slice(1).replace(/^[\\/]/, ""));
+  }
+  return path.resolve(process.cwd(), packageArg);
+}
+
+function formatPlugin(p: PluginRecord): string {
+  const statusColor =
+    p.status === "ready"
+      ? pc.green(p.status)
+      : p.status === "error"
+        ? pc.red(p.status)
+        : p.status === "disabled"
+          ? pc.dim(p.status)
+          : pc.yellow(p.status);
+
+  const parts = [
+    `key=${pc.bold(p.pluginKey)}`,
+    `status=${statusColor}`,
+    `version=${p.version}`,
+    `id=${pc.dim(p.id)}`,
+  ];
+
+  if (p.lastError) {
+    parts.push(`error=${pc.red(p.lastError.slice(0, 80))}`);
+  }
+
+  return parts.join("  ");
+}
+
+// ---------------------------------------------------------------------------
+// Command registration
+// ---------------------------------------------------------------------------
+
+export function registerPluginCommands(program: Command): void {
+  const plugin = program.command("plugin").description("Plugin lifecycle management");
+
+  // -------------------------------------------------------------------------
+  // plugin list
+  // -------------------------------------------------------------------------
+  addCommonClientOptions(
+    plugin
+      .command("list")
+      .description("List installed plugins")
+      .option("--status <status>", "Filter by status (ready, error, disabled, installed, upgrade_pending)")
+      .action(async (opts: PluginListOptions) => {
+        try {
+          const ctx = resolveCommandContext(opts);
+          const qs = opts.status ? `?status=${encodeURIComponent(opts.status)}` : "";
+          const plugins = await ctx.api.get<PluginRecord[]>(`/api/plugins${qs}`);
+
+          if (ctx.json) {
+            printOutput(plugins, { json: true });
+            return;
+          }
+
+          const rows = plugins ?? [];
+          if (rows.length === 0) {
+            console.log(pc.dim("No plugins installed."));
+            return;
+          }
+
+          for (const p of rows) {
+            console.log(formatPlugin(p));
+          }
+        } catch (err) {
+          handleCommandError(err);
+        }
+      }),
+  );
+
+  // -------------------------------------------------------------------------
+  // plugin install <package-or-path>
+  // -------------------------------------------------------------------------
+  addCommonClientOptions(
+    plugin
+      .command("install <package>")
+      .description(
+        "Install a plugin from a local path or npm package.\n" +
+          "  Examples:\n" +
+          "    paperclipai plugin install ./my-plugin              # local path\n" +
+          "    paperclipai plugin install @acme/plugin-linear      # npm package\n" +
+          "    paperclipai plugin install @acme/plugin-linear@1.2  # pinned version",
+      )
+      .option("-l, --local", "Treat <package> as a local filesystem path", false)
+      .option("--version <version>", "Specific npm version to install (npm packages only)")
+      .action(async (packageArg: string, opts: PluginInstallOptions) => {
+        try {
+          const ctx = resolveCommandContext(opts);
+
+          // Auto-detect local paths: starts with . or / or ~ or is an absolute path
+          const isLocal =
+            opts.local ||
+            packageArg.startsWith("./") ||
+            packageArg.startsWith("../") ||
+            packageArg.startsWith("/") ||
+            packageArg.startsWith("~");
+
+          const resolvedPackage = resolvePackageArg(packageArg, isLocal);
+
+          if (!ctx.json) {
+            console.log(
+              pc.dim(
+                isLocal
+                  ? `Installing plugin from local path: ${resolvedPackage}`
+                  : `Installing plugin: ${resolvedPackage}${opts.version ? `@${opts.version}` : ""}`,
+              ),
+            );
+          }
+
+          const installedPlugin = await ctx.api.post<PluginRecord>("/api/plugins/install", {
+            packageName: resolvedPackage,
+            version: opts.version,
+            isLocalPath: isLocal,
+          });
+
+          if (ctx.json) {
+            printOutput(installedPlugin, { json: true });
+            return;
+          }
+
+          if (!installedPlugin) {
+            console.log(pc.dim("Install returned no plugin record."));
+            return;
+          }
+
+          console.log(
+            pc.green(
+              `✓ Installed ${pc.bold(installedPlugin.pluginKey)} v${installedPlugin.version} (${installedPlugin.status})`,
+            ),
+          );
+
+          if (installedPlugin.lastError) {
+            console.log(pc.red(`  Warning: ${installedPlugin.lastError}`));
+          }
+        } catch (err) {
+          handleCommandError(err);
+        }
+      }),
+  );
+
+  // -------------------------------------------------------------------------
+  // plugin uninstall <plugin-key-or-id>
+  // -------------------------------------------------------------------------
+  addCommonClientOptions(
+    plugin
+      .command("uninstall <pluginKey>")
+      .description(
+        "Uninstall a plugin by its plugin key or database ID.\n" +
+          "  Use --force to hard-purge all state and config.",
+      )
+      .option("--force", "Purge all plugin state and config (hard delete)", false)
+      .action(async (pluginKey: string, opts: PluginUninstallOptions) => {
+        try {
+          const ctx = resolveCommandContext(opts);
+          const purge = opts.force === true;
+          const qs = purge ? "?purge=true" : "";
+
+          if (!ctx.json) {
+            console.log(
+              pc.dim(
+                purge
+                  ? `Uninstalling and purging plugin: ${pluginKey}`
+                  : `Uninstalling plugin: ${pluginKey}`,
+              ),
+            );
+          }
+
+          const result = await ctx.api.delete<PluginRecord | null>(
+            `/api/plugins/${encodeURIComponent(pluginKey)}${qs}`,
+          );
+
+          if (ctx.json) {
+            printOutput(result, { json: true });
+            return;
+          }
+
+          console.log(pc.green(`✓ Uninstalled ${pc.bold(pluginKey)}${purge ? " (purged)" : ""}`));
+        } catch (err) {
+          handleCommandError(err);
+        }
+      }),
+  );
+
+  // -------------------------------------------------------------------------
+  // plugin enable <plugin-key-or-id>
+  // -------------------------------------------------------------------------
+  addCommonClientOptions(
+    plugin
+      .command("enable <pluginKey>")
+      .description("Enable a disabled or errored plugin")
+      .action(async (pluginKey: string, opts: BaseClientOptions) => {
+        try {
+          const ctx = resolveCommandContext(opts);
+          const result = await ctx.api.post<PluginRecord>(
+            `/api/plugins/${encodeURIComponent(pluginKey)}/enable`,
+          );
+
+          if (ctx.json) {
+            printOutput(result, { json: true });
+            return;
+          }
+
+          console.log(pc.green(`✓ Enabled ${pc.bold(pluginKey)} — status: ${result?.status ?? "unknown"}`));
+        } catch (err) {
+          handleCommandError(err);
+        }
+      }),
+  );
+
+  // -------------------------------------------------------------------------
+  // plugin disable <plugin-key-or-id>
+  // -------------------------------------------------------------------------
+  addCommonClientOptions(
+    plugin
+      .command("disable <pluginKey>")
+      .description("Disable a running plugin without uninstalling it")
+      .action(async (pluginKey: string, opts: BaseClientOptions) => {
+        try {
+          const ctx = resolveCommandContext(opts);
+          const result = await ctx.api.post<PluginRecord>(
+            `/api/plugins/${encodeURIComponent(pluginKey)}/disable`,
+          );
+
+          if (ctx.json) {
+            printOutput(result, { json: true });
+            return;
+          }
+
+          console.log(pc.dim(`Disabled ${pc.bold(pluginKey)} — status: ${result?.status ?? "unknown"}`));
+        } catch (err) {
+          handleCommandError(err);
+        }
+      }),
+  );
+
+  // -------------------------------------------------------------------------
+  // plugin inspect <plugin-key-or-id>
+  // -------------------------------------------------------------------------
+  addCommonClientOptions(
+    plugin
+      .command("inspect <pluginKey>")
+      .description("Show full details for an installed plugin")
+      .action(async (pluginKey: string, opts: BaseClientOptions) => {
+        try {
+          const ctx = resolveCommandContext(opts);
+          const result = await ctx.api.get<PluginRecord>(
+            `/api/plugins/${encodeURIComponent(pluginKey)}`,
+          );
+
+          if (ctx.json) {
+            printOutput(result, { json: true });
+            return;
+          }
+
+          if (!result) {
+            console.log(pc.red(`Plugin not found: ${pluginKey}`));
+            process.exit(1);
+          }
+
+          console.log(formatPlugin(result));
+          if (result.lastError) {
+            console.log(`\n${pc.red("Last error:")}\n${result.lastError}`);
+          }
+        } catch (err) {
+          handleCommandError(err);
+        }
+      }),
+  );
+
+  // -------------------------------------------------------------------------
+  // plugin examples
+  // -------------------------------------------------------------------------
+  addCommonClientOptions(
+    plugin
+      .command("examples")
+      .description("List bundled example plugins available for local install")
+      .action(async (opts: BaseClientOptions) => {
+        try {
+          const ctx = resolveCommandContext(opts);
+          const examples = await ctx.api.get<
+            Array<{
+              packageName: string;
+              pluginKey: string;
+              displayName: string;
+              description: string;
+              localPath: string;
+              tag: string;
+            }>
+          >("/api/plugins/examples");
+
+          if (ctx.json) {
+            printOutput(examples, { json: true });
+            return;
+          }
+
+          const rows = examples ?? [];
+          if (rows.length === 0) {
+            console.log(pc.dim("No bundled examples available."));
+            return;
+          }
+
+          for (const ex of rows) {
+            console.log(
+              `${pc.bold(ex.displayName)}  ${pc.dim(ex.pluginKey)}\n` +
+                `  ${ex.description}\n` +
+                `  ${pc.cyan(`paperclipai plugin install ${ex.localPath}`)}`,
+            );
+          }
+        } catch (err) {
+          handleCommandError(err);
+        }
+      }),
+  );
+}

cli/src/commands/configure.ts

@@ -10,6 +10,7 @@ import { defaultSecretsConfig, promptSecrets } from "../prompts/secrets.js";
 import { defaultStorageConfig, promptStorage } from "../prompts/storage.js";
 import { promptServer } from "../prompts/server.js";
 import {
+  resolveDefaultBackupDir,
   resolveDefaultEmbeddedPostgresDir,
   resolveDefaultLogsDir,
   resolvePaperclipInstanceId,
@@ -39,6 +40,12 @@ function defaultConfig(): PaperclipConfig {
       mode: "embedded-postgres",
       embeddedPostgresDataDir: resolveDefaultEmbeddedPostgresDir(instanceId),
       embeddedPostgresPort: 54329,
+      backup: {
+        enabled: true,
+        intervalMinutes: 60,
+        retentionDays: 30,
+        dir: resolveDefaultBackupDir(instanceId),
+      },
     },
     logging: {
       mode: "file",
@@ -54,6 +61,7 @@ function defaultConfig(): PaperclipConfig {
     },
     auth: {
       baseUrlMode: "auto",
+      disableSignUp: false,
     },
     storage: defaultStorageConfig(),
     secrets: defaultSecretsConfig(),
@@ -118,7 +126,7 @@ export async function configure(opts: {
 
     switch (section) {
       case "database":
-        config.database = await promptDatabase();
+        config.database = await promptDatabase(config.database);
         break;
       case "llm": {
         const llm = await promptLlm();

cli/src/commands/db-backup.ts

@@ -0,0 +1,102 @@
+import path from "node:path";
+import * as p from "@clack/prompts";
+import pc from "picocolors";
+import { formatDatabaseBackupResult, runDatabaseBackup } from "@paperclipai/db";
+import {
+  expandHomePrefix,
+  resolveDefaultBackupDir,
+  resolvePaperclipInstanceId,
+} from "../config/home.js";
+import { readConfig, resolveConfigPath } from "../config/store.js";
+import { printPaperclipCliBanner } from "../utils/banner.js";
+
+type DbBackupOptions = {
+  config?: string;
+  dir?: string;
+  retentionDays?: number;
+  filenamePrefix?: string;
+  json?: boolean;
+};
+
+function resolveConnectionString(configPath?: string): { value: string; source: string } {
+  const envUrl = process.env.DATABASE_URL?.trim();
+  if (envUrl) return { value: envUrl, source: "DATABASE_URL" };
+
+  const config = readConfig(configPath);
+  if (config?.database.mode === "postgres" && config.database.connectionString?.trim()) {
+    return { value: config.database.connectionString.trim(), source: "config.database.connectionString" };
+  }
+
+  const port = config?.database.embeddedPostgresPort ?? 54329;
+  return {
+    value: `postgres://paperclip:paperclip@127.0.0.1:${port}/paperclip`,
+    source: `embedded-postgres@${port}`,
+  };
+}
+
+function normalizeRetentionDays(value: number | undefined, fallback: number): number {
+  const candidate = value ?? fallback;
+  if (!Number.isInteger(candidate) || candidate < 1) {
+    throw new Error(`Invalid retention days '${String(candidate)}'. Use a positive integer.`);
+  }
+  return candidate;
+}
+
+function resolveBackupDir(raw: string): string {
+  return path.resolve(expandHomePrefix(raw.trim()));
+}
+
+export async function dbBackupCommand(opts: DbBackupOptions): Promise<void> {
+  printPaperclipCliBanner();
+  p.intro(pc.bgCyan(pc.black(" paperclip db:backup ")));
+
+  const configPath = resolveConfigPath(opts.config);
+  const config = readConfig(opts.config);
+  const connection = resolveConnectionString(opts.config);
+  const defaultDir = resolveDefaultBackupDir(resolvePaperclipInstanceId());
+  const configuredDir = opts.dir?.trim() || config?.database.backup.dir || defaultDir;
+  const backupDir = resolveBackupDir(configuredDir);
+  const retentionDays = normalizeRetentionDays(
+    opts.retentionDays,
+    config?.database.backup.retentionDays ?? 30,
+  );
+  const filenamePrefix = opts.filenamePrefix?.trim() || "paperclip";
+
+  p.log.message(pc.dim(`Config: ${configPath}`));
+  p.log.message(pc.dim(`Connection source: ${connection.source}`));
+  p.log.message(pc.dim(`Backup dir: ${backupDir}`));
+  p.log.message(pc.dim(`Retention: ${retentionDays} day(s)`));
+
+  const spinner = p.spinner();
+  spinner.start("Creating database backup...");
+  try {
+    const result = await runDatabaseBackup({
+      connectionString: connection.value,
+      backupDir,
+      retentionDays,
+      filenamePrefix,
+    });
+    spinner.stop(`Backup saved: ${formatDatabaseBackupResult(result)}`);
+
+    if (opts.json) {
+      console.log(
+        JSON.stringify(
+          {
+            backupFile: result.backupFile,
+            sizeBytes: result.sizeBytes,
+            prunedCount: result.prunedCount,
+            backupDir,
+            retentionDays,
+            connectionSource: connection.source,
+          },
+          null,
+          2,
+        ),
+      );
+    }
+    p.outro(pc.green("Backup completed."));
+  } catch (err) {
+    spinner.stop(pc.red("Backup failed."));
+    throw err;
+  }
+}

cli/src/commands/doctor.ts

@@ -14,6 +14,7 @@ import {
   storageCheck,
   type CheckResult,
 } from "../checks/index.js";
+import { loadPaperclipEnvFile } from "../config/env.js";
 import { printPaperclipCliBanner } from "../utils/banner.js";
 
 const STATUS_ICON = {
@@ -31,6 +32,7 @@ export async function doctor(opts: {
   p.intro(pc.bgCyan(pc.black(" paperclip doctor ")));
 
   const configPath = resolveConfigPath(opts.config);
+  loadPaperclipEnvFile(configPath);
   const results: CheckResult[] = [];
 
   // 1. Config check (must pass before others)
@@ -64,28 +66,40 @@ export async function doctor(opts: {
   printResult(deploymentAuthResult);
 
   // 3. Agent JWT check
-  const jwtResult = agentJwtSecretCheck(opts.config);
-  results.push(jwtResult);
-  printResult(jwtResult);
-  await maybeRepair(jwtResult, opts);
+  results.push(
+    await runRepairableCheck({
+      run: () => agentJwtSecretCheck(opts.config),
+      configPath,
+      opts,
+    }),
+  );
 
   // 4. Secrets adapter check
-  const secretsResult = secretsCheck(config, configPath);
-  results.push(secretsResult);
-  printResult(secretsResult);
-  await maybeRepair(secretsResult, opts);
+  results.push(
+    await runRepairableCheck({
+      run: () => secretsCheck(config, configPath),
+      configPath,
+      opts,
+    }),
+  );
 
   // 5. Storage check
-  const storageResult = storageCheck(config, configPath);
-  results.push(storageResult);
-  printResult(storageResult);
-  await maybeRepair(storageResult, opts);
+  results.push(
+    await runRepairableCheck({
+      run: () => storageCheck(config, configPath),
+      configPath,
+      opts,
+    }),
+  );
 
   // 6. Database check
-  const dbResult = await databaseCheck(config, configPath);
-  results.push(dbResult);
-  printResult(dbResult);
-  await maybeRepair(dbResult, opts);
+  results.push(
+    await runRepairableCheck({
+      run: () => databaseCheck(config, configPath),
+      configPath,
+      opts,
+    }),
+  );
 
   // 7. LLM check
   const llmResult = await llmCheck(config);
@@ -93,10 +107,13 @@ export async function doctor(opts: {
   printResult(llmResult);
 
   // 8. Log directory check
-  const logResult = logCheck(config, configPath);
-  results.push(logResult);
-  printResult(logResult);
-  await maybeRepair(logResult, opts);
+  results.push(
+    await runRepairableCheck({
+      run: () => logCheck(config, configPath),
+      configPath,
+      opts,
+    }),
+  );
 
   // 9. Port check
   const portResult = await portCheck(config);
@@ -118,9 +135,9 @@ function printResult(result: CheckResult): void {
 async function maybeRepair(
   result: CheckResult,
   opts: { repair?: boolean; yes?: boolean },
-): Promise<void> {
-  if (result.status === "pass" || !result.canRepair || !result.repair) return;
-  if (!opts.repair) return;
+): Promise<boolean> {
+  if (result.status === "pass" || !result.canRepair || !result.repair) return false;
+  if (!opts.repair) return false;
 
   let shouldRepair = opts.yes;
   if (!shouldRepair) {
@@ -128,7 +145,7 @@ async function maybeRepair(
       message: `Repair "${result.name}"?`,
       initialValue: true,
     });
-    if (p.isCancel(answer)) return;
+    if (p.isCancel(answer)) return false;
     shouldRepair = answer;
   }
 
@@ -136,10 +153,30 @@ async function maybeRepair(
     try {
       await result.repair();
       p.log.success(`Repaired: ${result.name}`);
+      return true;
     } catch (err) {
       p.log.error(`Repair failed: ${err instanceof Error ? err.message : String(err)}`);
     }
   }
+  return false;
+}
+
+async function runRepairableCheck(input: {
+  run: () => CheckResult | Promise<CheckResult>;
+  configPath: string;
+  opts: { repair?: boolean; yes?: boolean };
+}): Promise<CheckResult> {
+  let result = await input.run();
+  printResult(result);
+
+  const repaired = await maybeRepair(result, input.opts);
+  if (!repaired) return result;
+
+  // Repairs may create/update the adjacent .env file or other local resources.
+  loadPaperclipEnvFile(input.configPath);
+  result = await input.run();
+  printResult(result);
+  return result;
 }
 
 function printSummary(results: CheckResult[]): { passed: number; warned: number; failed: number } {

cli/src/commands/env.ts

@@ -118,6 +118,29 @@ function collectDeploymentEnvRows(config: PaperclipConfig | null, configPath: st
   const dbUrl = process.env.DATABASE_URL ?? config?.database?.connectionString ?? "";
   const databaseMode = config?.database?.mode ?? "embedded-postgres";
   const dbUrlSource: EnvSource = process.env.DATABASE_URL ? "env" : config?.database?.connectionString ? "config" : "missing";
+  const publicUrl =
+    process.env.PAPERCLIP_PUBLIC_URL ??
+    process.env.PAPERCLIP_AUTH_PUBLIC_BASE_URL ??
+    process.env.BETTER_AUTH_URL ??
+    process.env.BETTER_AUTH_BASE_URL ??
+    config?.auth?.publicBaseUrl ??
+    "";
+  const publicUrlSource: EnvSource =
+    process.env.PAPERCLIP_PUBLIC_URL
+      ? "env"
+      : process.env.PAPERCLIP_AUTH_PUBLIC_BASE_URL || process.env.BETTER_AUTH_URL || process.env.BETTER_AUTH_BASE_URL
+        ? "env"
+        : config?.auth?.publicBaseUrl
+          ? "config"
+          : "missing";
+  let trustedOriginsDefault = "";
+  if (publicUrl) {
+    try {
+      trustedOriginsDefault = new URL(publicUrl).origin;
+    } catch {
+      trustedOriginsDefault = "";
+    }
+  }
 
   const heartbeatInterval = process.env.HEARTBEAT_SCHEDULER_INTERVAL_MS ?? DEFAULT_HEARTBEAT_SCHEDULER_INTERVAL_MS;
   const heartbeatEnabled = process.env.HEARTBEAT_SCHEDULER_ENABLED ?? "true";
@@ -192,6 +215,24 @@ function collectDeploymentEnvRows(config: PaperclipConfig | null, configPath: st
       required: false,
       note: "HTTP listen port",
     },
+    {
+      key: "PAPERCLIP_PUBLIC_URL",
+      value: publicUrl,
+      source: publicUrlSource,
+      required: false,
+      note: "Canonical public URL for auth/callback/invite origin wiring",
+    },
+    {
+      key: "BETTER_AUTH_TRUSTED_ORIGINS",
+      value: process.env.BETTER_AUTH_TRUSTED_ORIGINS ?? trustedOriginsDefault,
+      source: process.env.BETTER_AUTH_TRUSTED_ORIGINS
+        ? "env"
+        : trustedOriginsDefault
+          ? "default"
+          : "missing",
+      required: false,
+      note: "Comma-separated auth origin allowlist (auto-derived from PAPERCLIP_PUBLIC_URL when possible)",
+    },
     {
       key: "PAPERCLIP_AGENT_JWT_TTL_SECONDS",
       value: process.env.PAPERCLIP_AGENT_JWT_TTL_SECONDS ?? DEFAULT_AGENT_JWT_TTL_SECONDS,

cli/src/commands/onboard.ts

@@ -1,5 +1,18 @@
 import * as p from "@clack/prompts";
+import path from "node:path";
 import pc from "picocolors";
+import {
+  AUTH_BASE_URL_MODES,
+  DEPLOYMENT_EXPOSURES,
+  DEPLOYMENT_MODES,
+  SECRET_PROVIDERS,
+  STORAGE_PROVIDERS,
+  type AuthBaseUrlMode,
+  type DeploymentExposure,
+  type DeploymentMode,
+  type SecretProvider,
+  type StorageProvider,
+} from "@paperclipai/shared";
 import { configExists, readConfig, resolveConfigPath, writeConfig } from "../config/store.js";
 import type { PaperclipConfig } from "../config/schema.js";
 import { ensureAgentJwtSecret, resolveAgentJwtEnvFile } from "../config/env.js";
@@ -12,6 +25,8 @@ import { defaultStorageConfig, promptStorage } from "../prompts/storage.js";
 import { promptServer } from "../prompts/server.js";
 import {
   describeLocalInstancePaths,
+  expandHomePrefix,
+  resolveDefaultBackupDir,
   resolveDefaultEmbeddedPostgresDir,
   resolveDefaultLogsDir,
   resolvePaperclipInstanceId,
@@ -28,32 +43,194 @@ type OnboardOptions = {
   invokedByRun?: boolean;
 };
 
-function quickstartDefaults(): Pick<PaperclipConfig, "database" | "logging" | "server" | "auth" | "storage" | "secrets"> {
+type OnboardDefaults = Pick<PaperclipConfig, "database" | "logging" | "server" | "auth" | "storage" | "secrets">;
+
+const ONBOARD_ENV_KEYS = [
+  "PAPERCLIP_PUBLIC_URL",
+  "DATABASE_URL",
+  "PAPERCLIP_DB_BACKUP_ENABLED",
+  "PAPERCLIP_DB_BACKUP_INTERVAL_MINUTES",
+  "PAPERCLIP_DB_BACKUP_RETENTION_DAYS",
+  "PAPERCLIP_DB_BACKUP_DIR",
+  "PAPERCLIP_DEPLOYMENT_MODE",
+  "PAPERCLIP_DEPLOYMENT_EXPOSURE",
+  "HOST",
+  "PORT",
+  "SERVE_UI",
+  "PAPERCLIP_ALLOWED_HOSTNAMES",
+  "PAPERCLIP_AUTH_BASE_URL_MODE",
+  "PAPERCLIP_AUTH_PUBLIC_BASE_URL",
+  "BETTER_AUTH_URL",
+  "BETTER_AUTH_BASE_URL",
+  "PAPERCLIP_STORAGE_PROVIDER",
+  "PAPERCLIP_STORAGE_LOCAL_DIR",
+  "PAPERCLIP_STORAGE_S3_BUCKET",
+  "PAPERCLIP_STORAGE_S3_REGION",
+  "PAPERCLIP_STORAGE_S3_ENDPOINT",
+  "PAPERCLIP_STORAGE_S3_PREFIX",
+  "PAPERCLIP_STORAGE_S3_FORCE_PATH_STYLE",
+  "PAPERCLIP_SECRETS_PROVIDER",
+  "PAPERCLIP_SECRETS_STRICT_MODE",
+  "PAPERCLIP_SECRETS_MASTER_KEY_FILE",
+] as const;
+
+function parseBooleanFromEnv(rawValue: string | undefined): boolean | null {
+  if (rawValue === undefined) return null;
+  const lower = rawValue.trim().toLowerCase();
+  if (lower === "true" || lower === "1" || lower === "yes") return true;
+  if (lower === "false" || lower === "0" || lower === "no") return false;
+  return null;
+}
+
+function parseNumberFromEnv(rawValue: string | undefined): number | null {
+  if (!rawValue) return null;
+  const parsed = Number(rawValue);
+  if (!Number.isFinite(parsed)) return null;
+  return parsed;
+}
+
+function parseEnumFromEnv<T extends string>(rawValue: string | undefined, allowedValues: readonly T[]): T | null {
+  if (!rawValue) return null;
+  return allowedValues.includes(rawValue as T) ? (rawValue as T) : null;
+}
+
+function resolvePathFromEnv(rawValue: string | undefined): string | null {
+  if (!rawValue || rawValue.trim().length === 0) return null;
+  return path.resolve(expandHomePrefix(rawValue.trim()));
+}
+
+function quickstartDefaultsFromEnv(): {
+  defaults: OnboardDefaults;
+  usedEnvKeys: string[];
+  ignoredEnvKeys: Array<{ key: string; reason: string }>;
+} {
   const instanceId = resolvePaperclipInstanceId();
-  return {
+  const defaultStorage = defaultStorageConfig();
+  const defaultSecrets = defaultSecretsConfig();
+  const databaseUrl = process.env.DATABASE_URL?.trim() || undefined;
+  const publicUrl =
+    process.env.PAPERCLIP_PUBLIC_URL?.trim() ||
+    process.env.PAPERCLIP_AUTH_PUBLIC_BASE_URL?.trim() ||
+    process.env.BETTER_AUTH_URL?.trim() ||
+    process.env.BETTER_AUTH_BASE_URL?.trim() ||
+    undefined;
+  const deploymentMode =
+    parseEnumFromEnv<DeploymentMode>(process.env.PAPERCLIP_DEPLOYMENT_MODE, DEPLOYMENT_MODES) ?? "local_trusted";
+  const deploymentExposureFromEnv = parseEnumFromEnv<DeploymentExposure>(
+    process.env.PAPERCLIP_DEPLOYMENT_EXPOSURE,
+    DEPLOYMENT_EXPOSURES,
+  );
+  const deploymentExposure =
+    deploymentMode === "local_trusted" ? "private" : (deploymentExposureFromEnv ?? "private");
+  const authPublicBaseUrl = publicUrl;
+  const authBaseUrlModeFromEnv = parseEnumFromEnv<AuthBaseUrlMode>(
+    process.env.PAPERCLIP_AUTH_BASE_URL_MODE,
+    AUTH_BASE_URL_MODES,
+  );
+  const authBaseUrlMode = authBaseUrlModeFromEnv ?? (authPublicBaseUrl ? "explicit" : "auto");
+  const allowedHostnamesFromEnv = process.env.PAPERCLIP_ALLOWED_HOSTNAMES
+    ? process.env.PAPERCLIP_ALLOWED_HOSTNAMES
+      .split(",")
+      .map((value) => value.trim().toLowerCase())
+      .filter((value) => value.length > 0)
+    : [];
+  const hostnameFromPublicUrl = publicUrl
+    ? (() => {
+      try {
+        return new URL(publicUrl).hostname.trim().toLowerCase();
+      } catch {
+        return null;
+      }
+    })()
+    : null;
+  const storageProvider =
+    parseEnumFromEnv<StorageProvider>(process.env.PAPERCLIP_STORAGE_PROVIDER, STORAGE_PROVIDERS) ??
+    defaultStorage.provider;
+  const secretsProvider =
+    parseEnumFromEnv<SecretProvider>(process.env.PAPERCLIP_SECRETS_PROVIDER, SECRET_PROVIDERS) ??
+    defaultSecrets.provider;
+  const databaseBackupEnabled = parseBooleanFromEnv(process.env.PAPERCLIP_DB_BACKUP_ENABLED) ?? true;
+  const databaseBackupIntervalMinutes = Math.max(
+    1,
+    parseNumberFromEnv(process.env.PAPERCLIP_DB_BACKUP_INTERVAL_MINUTES) ?? 60,
+  );
+  const databaseBackupRetentionDays = Math.max(
+    1,
+    parseNumberFromEnv(process.env.PAPERCLIP_DB_BACKUP_RETENTION_DAYS) ?? 30,
+  );
+  const defaults: OnboardDefaults = {
     database: {
-      mode: "embedded-postgres",
+      mode: databaseUrl ? "postgres" : "embedded-postgres",
+      ...(databaseUrl ? { connectionString: databaseUrl } : {}),
       embeddedPostgresDataDir: resolveDefaultEmbeddedPostgresDir(instanceId),
       embeddedPostgresPort: 54329,
+      backup: {
+        enabled: databaseBackupEnabled,
+        intervalMinutes: databaseBackupIntervalMinutes,
+        retentionDays: databaseBackupRetentionDays,
+        dir: resolvePathFromEnv(process.env.PAPERCLIP_DB_BACKUP_DIR) ?? resolveDefaultBackupDir(instanceId),
+      },
     },
     logging: {
       mode: "file",
       logDir: resolveDefaultLogsDir(instanceId),
     },
     server: {
-      deploymentMode: "local_trusted",
-      exposure: "private",
-      host: "127.0.0.1",
-      port: 3100,
-      allowedHostnames: [],
-      serveUi: true,
+      deploymentMode,
+      exposure: deploymentExposure,
+      host: process.env.HOST ?? "127.0.0.1",
+      port: Number(process.env.PORT) || 3100,
+      allowedHostnames: Array.from(new Set([...allowedHostnamesFromEnv, ...(hostnameFromPublicUrl ? [hostnameFromPublicUrl] : [])])),
+      serveUi: parseBooleanFromEnv(process.env.SERVE_UI) ?? true,
     },
     auth: {
-      baseUrlMode: "auto",
+      baseUrlMode: authBaseUrlMode,
+      disableSignUp: false,
+      ...(authPublicBaseUrl ? { publicBaseUrl: authPublicBaseUrl } : {}),
+    },
+    storage: {
+      provider: storageProvider,
+      localDisk: {
+        baseDir:
+          resolvePathFromEnv(process.env.PAPERCLIP_STORAGE_LOCAL_DIR) ?? defaultStorage.localDisk.baseDir,
+      },
+      s3: {
+        bucket: process.env.PAPERCLIP_STORAGE_S3_BUCKET ?? defaultStorage.s3.bucket,
+        region: process.env.PAPERCLIP_STORAGE_S3_REGION ?? defaultStorage.s3.region,
+        endpoint: process.env.PAPERCLIP_STORAGE_S3_ENDPOINT ?? defaultStorage.s3.endpoint,
+        prefix: process.env.PAPERCLIP_STORAGE_S3_PREFIX ?? defaultStorage.s3.prefix,
+        forcePathStyle:
+          parseBooleanFromEnv(process.env.PAPERCLIP_STORAGE_S3_FORCE_PATH_STYLE) ??
+          defaultStorage.s3.forcePathStyle,
+      },
+    },
+    secrets: {
+      provider: secretsProvider,
+      strictMode: parseBooleanFromEnv(process.env.PAPERCLIP_SECRETS_STRICT_MODE) ?? defaultSecrets.strictMode,
+      localEncrypted: {
+        keyFilePath:
+          resolvePathFromEnv(process.env.PAPERCLIP_SECRETS_MASTER_KEY_FILE) ??
+          defaultSecrets.localEncrypted.keyFilePath,
+      },
     },
-    storage: defaultStorageConfig(),
-    secrets: defaultSecretsConfig(),
   };
+  const ignoredEnvKeys: Array<{ key: string; reason: string }> = [];
+  if (deploymentMode === "local_trusted" && process.env.PAPERCLIP_DEPLOYMENT_EXPOSURE !== undefined) {
+    ignoredEnvKeys.push({
+      key: "PAPERCLIP_DEPLOYMENT_EXPOSURE",
+      reason: "Ignored because deployment mode local_trusted always forces private exposure",
+    });
+  }
+
+  const ignoredKeySet = new Set(ignoredEnvKeys.map((entry) => entry.key));
+  const usedEnvKeys = ONBOARD_ENV_KEYS.filter(
+    (key) => process.env[key] !== undefined && !ignoredKeySet.has(key),
+  );
+  return { defaults, usedEnvKeys, ignoredEnvKeys };
+}
+
+function canCreateBootstrapInviteImmediately(config: Pick<PaperclipConfig, "database" | "server">): boolean {
+  return config.server.deploymentMode === "authenticated" && config.database.mode !== "embedded-postgres";
 }
 
 export async function onboard(opts: OnboardOptions): Promise<void> {
@@ -109,6 +286,7 @@ export async function onboard(opts: OnboardOptions): Promise<void> {
   }
 
   let llm: PaperclipConfig["llm"] | undefined;
+  const { defaults: derivedDefaults, usedEnvKeys, ignoredEnvKeys } = quickstartDefaultsFromEnv();
   let {
     database,
     logging,
@@ -116,11 +294,11 @@ export async function onboard(opts: OnboardOptions): Promise<void> {
     auth,
     storage,
     secrets,
-  } = quickstartDefaults();
+  } = derivedDefaults;
 
   if (setupMode === "advanced") {
     p.log.step(pc.bold("Database"));
-    database = await promptDatabase();
+    database = await promptDatabase(database);
 
     if (database.mode === "postgres" && database.connectionString) {
       const s = p.spinner();
@@ -184,13 +362,20 @@ export async function onboard(opts: OnboardOptions): Promise<void> {
     logging = await promptLogging();
 
     p.log.step(pc.bold("Server"));
-    ({ server, auth } = await promptServer());
+    ({ server, auth } = await promptServer({ currentServer: server, currentAuth: auth }));
 
     p.log.step(pc.bold("Storage"));
-    storage = await promptStorage(defaultStorageConfig());
+    storage = await promptStorage(storage);
 
     p.log.step(pc.bold("Secrets"));
-    secrets = defaultSecretsConfig();
+    const secretsDefaults = defaultSecretsConfig();
+    secrets = {
+      provider: secrets.provider ?? secretsDefaults.provider,
+      strictMode: secrets.strictMode ?? secretsDefaults.strictMode,
+      localEncrypted: {
+        keyFilePath: secrets.localEncrypted?.keyFilePath ?? secretsDefaults.localEncrypted.keyFilePath,
+      },
+    };
     p.log.message(
       pc.dim(
         `Using defaults: provider=${secrets.provider}, strictMode=${secrets.strictMode}, keyFile=${secrets.localEncrypted.keyFilePath}`,
@@ -198,9 +383,17 @@ export async function onboard(opts: OnboardOptions): Promise<void> {
     );
   } else {
     p.log.step(pc.bold("Quickstart"));
-    p.log.message(
-      pc.dim("Using local defaults: embedded database, no LLM provider, file storage, and local encrypted secrets."),
-    );
+    p.log.message(pc.dim("Using quickstart defaults."));
+    if (usedEnvKeys.length > 0) {
+      p.log.message(pc.dim(`Environment-aware defaults active (${usedEnvKeys.length} env var(s) detected).`));
+    } else {
+      p.log.message(
+        pc.dim("No environment overrides detected: embedded database, file storage, local encrypted secrets."),
+      );
+    }
+    for (const ignored of ignoredEnvKeys) {
+      p.log.message(pc.dim(`Ignored ${ignored.key}: ${ignored.reason}`));
+    }
   }
 
   const jwtSecret = ensureAgentJwtSecret(configPath);
@@ -261,7 +454,7 @@ export async function onboard(opts: OnboardOptions): Promise<void> {
     "Next commands",
   );
 
-  if (server.deploymentMode === "authenticated") {
+  if (canCreateBootstrapInviteImmediately({ database, server })) {
     p.log.step("Generating bootstrap CEO invite");
     await bootstrapCeoInvite({ config: configPath });
   }
@@ -284,5 +477,15 @@ export async function onboard(opts: OnboardOptions): Promise<void> {
     return;
   }
 
+  if (server.deploymentMode === "authenticated" && database.mode === "embedded-postgres") {
+    p.log.info(
+      [
+        "Bootstrap CEO invite will be created after the server starts.",
+        `Next: ${pc.cyan("paperclipai run")}`,
+        `Then: ${pc.cyan("paperclipai auth bootstrap-ceo")}`,
+      ].join("\n"),
+    );
+  }
+
   p.outro("You're all set!");
 }

cli/src/commands/run.ts

@@ -3,9 +3,13 @@ import path from "node:path";
 import { fileURLToPath, pathToFileURL } from "node:url";
 import * as p from "@clack/prompts";
 import pc from "picocolors";
+import { bootstrapCeoInvite } from "./auth-bootstrap-ceo.js";
 import { onboard } from "./onboard.js";
 import { doctor } from "./doctor.js";
+import { loadPaperclipEnvFile } from "../config/env.js";
 import { configExists, resolveConfigPath } from "../config/store.js";
+import type { PaperclipConfig } from "../config/schema.js";
+import { readConfig } from "../config/store.js";
 import {
   describeLocalInstancePaths,
   resolvePaperclipHomeDir,
@@ -19,6 +23,13 @@ interface RunOptions {
   yes?: boolean;
 }
 
+interface StartedServer {
+  apiUrl: string;
+  databaseUrl: string;
+  host: string;
+  listenPort: number;
+}
+
 export async function runCommand(opts: RunOptions): Promise<void> {
   const instanceId = resolvePaperclipInstanceId(opts.instance);
   process.env.PAPERCLIP_INSTANCE_ID = instanceId;
@@ -31,6 +42,7 @@ export async function runCommand(opts: RunOptions): Promise<void> {
 
   const configPath = resolveConfigPath(opts.config);
   process.env.PAPERCLIP_CONFIG = configPath;
+  loadPaperclipEnvFile(configPath);
 
   p.intro(pc.bgCyan(pc.black(" paperclipai run ")));
   p.log.message(pc.dim(`Home: ${paths.homeDir}`));
@@ -60,8 +72,41 @@ export async function runCommand(opts: RunOptions): Promise<void> {
     process.exit(1);
   }
 
+  const config = readConfig(configPath);
+  if (!config) {
+    p.log.error(`No config found at ${configPath}.`);
+    process.exit(1);
+  }
+
   p.log.step("Starting Paperclip server...");
-  await importServerEntry();
+  const startedServer = await importServerEntry();
+
+  if (shouldGenerateBootstrapInviteAfterStart(config)) {
+    p.log.step("Generating bootstrap CEO invite");
+    await bootstrapCeoInvite({
+      config: configPath,
+      dbUrl: startedServer.databaseUrl,
+      baseUrl: resolveBootstrapInviteBaseUrl(config, startedServer),
+    });
+  }
+}
+
+function resolveBootstrapInviteBaseUrl(
+  config: PaperclipConfig,
+  startedServer: StartedServer,
+): string {
+  const explicitBaseUrl =
+    process.env.PAPERCLIP_PUBLIC_URL ??
+    process.env.PAPERCLIP_AUTH_PUBLIC_BASE_URL ??
+    process.env.BETTER_AUTH_URL ??
+    process.env.BETTER_AUTH_BASE_URL ??
+    (config.auth.baseUrlMode === "explicit" ? config.auth.publicBaseUrl : undefined);
+
+  if (typeof explicitBaseUrl === "string" && explicitBaseUrl.trim().length > 0) {
+    return explicitBaseUrl.trim().replace(/\/+$/, "");
+  }
+
+  return startedServer.apiUrl.replace(/\/api$/, "");
 }
 
 function formatError(err: unknown): string {
@@ -101,19 +146,20 @@ function maybeEnableUiDevMiddleware(entrypoint: string): void {
   }
 }
 
-async function importServerEntry(): Promise<void> {
+async function importServerEntry(): Promise<StartedServer> {
   // Dev mode: try local workspace path (monorepo with tsx)
   const projectRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../../..");
   const devEntry = path.resolve(projectRoot, "server/src/index.ts");
   if (fs.existsSync(devEntry)) {
     maybeEnableUiDevMiddleware(devEntry);
-    await import(pathToFileURL(devEntry).href);
-    return;
+    const mod = await import(pathToFileURL(devEntry).href);
+    return await startServerFromModule(mod, devEntry);
   }
 
   // Production mode: import the published @paperclipai/server package
   try {
-    await import("@paperclipai/server");
+    const mod = await import("@paperclipai/server");
+    return await startServerFromModule(mod, "@paperclipai/server");
   } catch (err) {
     const missingSpecifier = getMissingModuleSpecifier(err);
     const missingServerEntrypoint = !missingSpecifier || missingSpecifier === "@paperclipai/server";
@@ -130,3 +176,15 @@ async function importServerEntry(): Promise<void> {
     );
   }
 }
+
+function shouldGenerateBootstrapInviteAfterStart(config: PaperclipConfig): boolean {
+  return config.server.deploymentMode === "authenticated" && config.database.mode === "embedded-postgres";
+}
+
+async function startServerFromModule(mod: unknown, label: string): Promise<StartedServer> {
+  const startServer = (mod as { startServer?: () => Promise<StartedServer> }).startServer;
+  if (typeof startServer !== "function") {
+    throw new Error(`Paperclip server entrypoint did not export startServer(): ${label}`);
+  }
+  return await startServer();
+}

cli/src/commands/worktree-lib.ts

@@ -0,0 +1,274 @@
+import { randomInt } from "node:crypto";
+import path from "node:path";
+import type { PaperclipConfig } from "../config/schema.js";
+import { expandHomePrefix } from "../config/home.js";
+
+export const DEFAULT_WORKTREE_HOME = "~/.paperclip-worktrees";
+export const WORKTREE_SEED_MODES = ["minimal", "full"] as const;
+
+export type WorktreeSeedMode = (typeof WORKTREE_SEED_MODES)[number];
+
+export type WorktreeSeedPlan = {
+  mode: WorktreeSeedMode;
+  excludedTables: string[];
+  nullifyColumns: Record<string, string[]>;
+};
+
+const MINIMAL_WORKTREE_EXCLUDED_TABLES = [
+  "activity_log",
+  "agent_runtime_state",
+  "agent_task_sessions",
+  "agent_wakeup_requests",
+  "cost_events",
+  "heartbeat_run_events",
+  "heartbeat_runs",
+  "workspace_runtime_services",
+];
+
+const MINIMAL_WORKTREE_NULLIFIED_COLUMNS: Record<string, string[]> = {
+  issues: ["checkout_run_id", "execution_run_id"],
+};
+
+export type WorktreeLocalPaths = {
+  cwd: string;
+  repoConfigDir: string;
+  configPath: string;
+  envPath: string;
+  homeDir: string;
+  instanceId: string;
+  instanceRoot: string;
+  contextPath: string;
+  embeddedPostgresDataDir: string;
+  backupDir: string;
+  logDir: string;
+  secretsKeyFilePath: string;
+  storageDir: string;
+};
+
+export type WorktreeUiBranding = {
+  name: string;
+  color: string;
+};
+
+export function isWorktreeSeedMode(value: string): value is WorktreeSeedMode {
+  return (WORKTREE_SEED_MODES as readonly string[]).includes(value);
+}
+
+export function resolveWorktreeSeedPlan(mode: WorktreeSeedMode): WorktreeSeedPlan {
+  if (mode === "full") {
+    return {
+      mode,
+      excludedTables: [],
+      nullifyColumns: {},
+    };
+  }
+  return {
+    mode,
+    excludedTables: [...MINIMAL_WORKTREE_EXCLUDED_TABLES],
+    nullifyColumns: {
+      ...MINIMAL_WORKTREE_NULLIFIED_COLUMNS,
+    },
+  };
+}
+
+function nonEmpty(value: string | null | undefined): string | null {
+  return typeof value === "string" && value.trim().length > 0 ? value.trim() : null;
+}
+
+function isLoopbackHost(hostname: string): boolean {
+  const value = hostname.trim().toLowerCase();
+  return value === "127.0.0.1" || value === "localhost" || value === "::1";
+}
+
+export function sanitizeWorktreeInstanceId(rawValue: string): string {
+  const trimmed = rawValue.trim().toLowerCase();
+  const normalized = trimmed
+    .replace(/[^a-z0-9_-]+/g, "-")
+    .replace(/-+/g, "-")
+    .replace(/^[-_]+|[-_]+$/g, "");
+  return normalized || "worktree";
+}
+
+export function resolveSuggestedWorktreeName(cwd: string, explicitName?: string): string {
+  return nonEmpty(explicitName) ?? path.basename(path.resolve(cwd));
+}
+
+function hslComponentToHex(n: number): string {
+  return Math.round(Math.max(0, Math.min(255, n)))
+    .toString(16)
+    .padStart(2, "0");
+}
+
+function hslToHex(hue: number, saturation: number, lightness: number): string {
+  const s = Math.max(0, Math.min(100, saturation)) / 100;
+  const l = Math.max(0, Math.min(100, lightness)) / 100;
+  const c = (1 - Math.abs((2 * l) - 1)) * s;
+  const h = ((hue % 360) + 360) % 360;
+  const x = c * (1 - Math.abs(((h / 60) % 2) - 1));
+  const m = l - (c / 2);
+
+  let r = 0;
+  let g = 0;
+  let b = 0;
+
+  if (h < 60) {
+    r = c;
+    g = x;
+  } else if (h < 120) {
+    r = x;
+    g = c;
+  } else if (h < 180) {
+    g = c;
+    b = x;
+  } else if (h < 240) {
+    g = x;
+    b = c;
+  } else if (h < 300) {
+    r = x;
+    b = c;
+  } else {
+    r = c;
+    b = x;
+  }
+
+  return `#${hslComponentToHex((r + m) * 255)}${hslComponentToHex((g + m) * 255)}${hslComponentToHex((b + m) * 255)}`;
+}
+
+export function generateWorktreeColor(): string {
+  return hslToHex(randomInt(0, 360), 68, 56);
+}
+
+export function resolveWorktreeLocalPaths(opts: {
+  cwd: string;
+  homeDir?: string;
+  instanceId: string;
+}): WorktreeLocalPaths {
+  const cwd = path.resolve(opts.cwd);
+  const homeDir = path.resolve(expandHomePrefix(opts.homeDir ?? DEFAULT_WORKTREE_HOME));
+  const instanceRoot = path.resolve(homeDir, "instances", opts.instanceId);
+  const repoConfigDir = path.resolve(cwd, ".paperclip");
+  return {
+    cwd,
+    repoConfigDir,
+    configPath: path.resolve(repoConfigDir, "config.json"),
+    envPath: path.resolve(repoConfigDir, ".env"),
+    homeDir,
+    instanceId: opts.instanceId,
+    instanceRoot,
+    contextPath: path.resolve(homeDir, "context.json"),
+    embeddedPostgresDataDir: path.resolve(instanceRoot, "db"),
+    backupDir: path.resolve(instanceRoot, "data", "backups"),
+    logDir: path.resolve(instanceRoot, "logs"),
+    secretsKeyFilePath: path.resolve(instanceRoot, "secrets", "master.key"),
+    storageDir: path.resolve(instanceRoot, "data", "storage"),
+  };
+}
+
+export function rewriteLocalUrlPort(rawUrl: string | undefined, port: number): string | undefined {
+  if (!rawUrl) return undefined;
+  try {
+    const parsed = new URL(rawUrl);
+    if (!isLoopbackHost(parsed.hostname)) return rawUrl;
+    parsed.port = String(port);
+    return parsed.toString();
+  } catch {
+    return rawUrl;
+  }
+}
+
+export function buildWorktreeConfig(input: {
+  sourceConfig: PaperclipConfig | null;
+  paths: WorktreeLocalPaths;
+  serverPort: number;
+  databasePort: number;
+  now?: Date;
+}): PaperclipConfig {
+  const { sourceConfig, paths, serverPort, databasePort } = input;
+  const nowIso = (input.now ?? new Date()).toISOString();
+
+  const source = sourceConfig;
+  const authPublicBaseUrl = rewriteLocalUrlPort(source?.auth.publicBaseUrl, serverPort);
+
+  return {
+    $meta: {
+      version: 1,
+      updatedAt: nowIso,
+      source: "configure",
+    },
+    ...(source?.llm ? { llm: source.llm } : {}),
+    database: {
+      mode: "embedded-postgres",
+      embeddedPostgresDataDir: paths.embeddedPostgresDataDir,
+      embeddedPostgresPort: databasePort,
+      backup: {
+        enabled: source?.database.backup.enabled ?? true,
+        intervalMinutes: source?.database.backup.intervalMinutes ?? 60,
+        retentionDays: source?.database.backup.retentionDays ?? 30,
+        dir: paths.backupDir,
+      },
+    },
+    logging: {
+      mode: source?.logging.mode ?? "file",
+      logDir: paths.logDir,
+    },
+    server: {
+      deploymentMode: source?.server.deploymentMode ?? "local_trusted",
+      exposure: source?.server.exposure ?? "private",
+      host: source?.server.host ?? "127.0.0.1",
+      port: serverPort,
+      allowedHostnames: source?.server.allowedHostnames ?? [],
+      serveUi: source?.server.serveUi ?? true,
+    },
+    auth: {
+      baseUrlMode: source?.auth.baseUrlMode ?? "auto",
+      ...(authPublicBaseUrl ? { publicBaseUrl: authPublicBaseUrl } : {}),
+      disableSignUp: source?.auth.disableSignUp ?? false,
+    },
+    storage: {
+      provider: source?.storage.provider ?? "local_disk",
+      localDisk: {
+        baseDir: paths.storageDir,
+      },
+      s3: {
+        bucket: source?.storage.s3.bucket ?? "paperclip",
+        region: source?.storage.s3.region ?? "us-east-1",
+        endpoint: source?.storage.s3.endpoint,
+        prefix: source?.storage.s3.prefix ?? "",
+        forcePathStyle: source?.storage.s3.forcePathStyle ?? false,
+      },
+    },
+    secrets: {
+      provider: source?.secrets.provider ?? "local_encrypted",
+      strictMode: source?.secrets.strictMode ?? false,
+      localEncrypted: {
+        keyFilePath: paths.secretsKeyFilePath,
+      },
+    },
+  };
+}
+
+export function buildWorktreeEnvEntries(
+  paths: WorktreeLocalPaths,
+  branding?: WorktreeUiBranding,
+): Record<string, string> {
+  return {
+    PAPERCLIP_HOME: paths.homeDir,
+    PAPERCLIP_INSTANCE_ID: paths.instanceId,
+    PAPERCLIP_CONFIG: paths.configPath,
+    PAPERCLIP_CONTEXT: paths.contextPath,
+    PAPERCLIP_IN_WORKTREE: "true",
+    ...(branding?.name ? { PAPERCLIP_WORKTREE_NAME: branding.name } : {}),
+    ...(branding?.color ? { PAPERCLIP_WORKTREE_COLOR: branding.color } : {}),
+  };
+}
+
+function shellEscape(value: string): string {
+  return `'${value.replaceAll("'", `'\"'\"'`)}'`;
+}
+
+export function formatShellExports(entries: Record<string, string>): string {
+  return Object.entries(entries)
+    .filter(([, value]) => typeof value === "string" && value.trim().length > 0)
+    .map(([key, value]) => `export ${key}=${shellEscape(value)}`)
+    .join("\n");
+}

cli/src/commands/worktree-merge-history-lib.ts

@@ -0,0 +1,709 @@
+import {
+  agents,
+  assets,
+  documentRevisions,
+  goals,
+  issueAttachments,
+  issueComments,
+  issueDocuments,
+  issues,
+  projects,
+  projectWorkspaces,
+} from "@paperclipai/db";
+
+type IssueRow = typeof issues.$inferSelect;
+type CommentRow = typeof issueComments.$inferSelect;
+type AgentRow = typeof agents.$inferSelect;
+type ProjectRow = typeof projects.$inferSelect;
+type ProjectWorkspaceRow = typeof projectWorkspaces.$inferSelect;
+type GoalRow = typeof goals.$inferSelect;
+type IssueDocumentLinkRow = typeof issueDocuments.$inferSelect;
+type DocumentRevisionTableRow = typeof documentRevisions.$inferSelect;
+type IssueAttachmentTableRow = typeof issueAttachments.$inferSelect;
+type AssetRow = typeof assets.$inferSelect;
+
+export const WORKTREE_MERGE_SCOPES = ["issues", "comments"] as const;
+export type WorktreeMergeScope = (typeof WORKTREE_MERGE_SCOPES)[number];
+
+export type ImportAdjustment =
+  | "clear_assignee_agent"
+  | "clear_project"
+  | "clear_project_workspace"
+  | "clear_goal"
+  | "clear_author_agent"
+  | "coerce_in_progress_to_todo"
+  | "clear_document_agent"
+  | "clear_document_revision_agent"
+  | "clear_attachment_agent";
+
+export type IssueMergeAction = "skip_existing" | "insert";
+export type CommentMergeAction = "skip_existing" | "skip_missing_parent" | "insert";
+
+export type PlannedIssueInsert = {
+  source: IssueRow;
+  action: "insert";
+  previewIssueNumber: number;
+  previewIdentifier: string;
+  targetStatus: string;
+  targetAssigneeAgentId: string | null;
+  targetCreatedByAgentId: string | null;
+  targetProjectId: string | null;
+  targetProjectWorkspaceId: string | null;
+  targetGoalId: string | null;
+  projectResolution: "preserved" | "cleared" | "mapped";
+  mappedProjectName: string | null;
+  adjustments: ImportAdjustment[];
+};
+
+export type PlannedIssueSkip = {
+  source: IssueRow;
+  action: "skip_existing";
+  driftKeys: string[];
+};
+
+export type PlannedCommentInsert = {
+  source: CommentRow;
+  action: "insert";
+  targetAuthorAgentId: string | null;
+  adjustments: ImportAdjustment[];
+};
+
+export type PlannedCommentSkip = {
+  source: CommentRow;
+  action: "skip_existing" | "skip_missing_parent";
+};
+
+export type IssueDocumentRow = {
+  id: IssueDocumentLinkRow["id"];
+  companyId: IssueDocumentLinkRow["companyId"];
+  issueId: IssueDocumentLinkRow["issueId"];
+  documentId: IssueDocumentLinkRow["documentId"];
+  key: IssueDocumentLinkRow["key"];
+  linkCreatedAt: IssueDocumentLinkRow["createdAt"];
+  linkUpdatedAt: IssueDocumentLinkRow["updatedAt"];
+  title: string | null;
+  format: string;
+  latestBody: string;
+  latestRevisionId: string | null;
+  latestRevisionNumber: number;
+  createdByAgentId: string | null;
+  createdByUserId: string | null;
+  updatedByAgentId: string | null;
+  updatedByUserId: string | null;
+  documentCreatedAt: Date;
+  documentUpdatedAt: Date;
+};
+
+export type DocumentRevisionRow = {
+  id: DocumentRevisionTableRow["id"];
+  companyId: DocumentRevisionTableRow["companyId"];
+  documentId: DocumentRevisionTableRow["documentId"];
+  revisionNumber: DocumentRevisionTableRow["revisionNumber"];
+  body: DocumentRevisionTableRow["body"];
+  changeSummary: DocumentRevisionTableRow["changeSummary"];
+  createdByAgentId: string | null;
+  createdByUserId: string | null;
+  createdAt: Date;
+};
+
+export type IssueAttachmentRow = {
+  id: IssueAttachmentTableRow["id"];
+  companyId: IssueAttachmentTableRow["companyId"];
+  issueId: IssueAttachmentTableRow["issueId"];
+  issueCommentId: IssueAttachmentTableRow["issueCommentId"];
+  assetId: IssueAttachmentTableRow["assetId"];
+  provider: AssetRow["provider"];
+  objectKey: AssetRow["objectKey"];
+  contentType: AssetRow["contentType"];
+  byteSize: AssetRow["byteSize"];
+  sha256: AssetRow["sha256"];
+  originalFilename: AssetRow["originalFilename"];
+  createdByAgentId: string | null;
+  createdByUserId: string | null;
+  assetCreatedAt: Date;
+  assetUpdatedAt: Date;
+  attachmentCreatedAt: Date;
+  attachmentUpdatedAt: Date;
+};
+
+export type PlannedDocumentRevisionInsert = {
+  source: DocumentRevisionRow;
+  targetRevisionNumber: number;
+  targetCreatedByAgentId: string | null;
+  adjustments: ImportAdjustment[];
+};
+
+export type PlannedIssueDocumentInsert = {
+  source: IssueDocumentRow;
+  action: "insert";
+  targetCreatedByAgentId: string | null;
+  targetUpdatedByAgentId: string | null;
+  latestRevisionId: string | null;
+  latestRevisionNumber: number;
+  revisionsToInsert: PlannedDocumentRevisionInsert[];
+  adjustments: ImportAdjustment[];
+};
+
+export type PlannedIssueDocumentMerge = {
+  source: IssueDocumentRow;
+  action: "merge_existing";
+  targetCreatedByAgentId: string | null;
+  targetUpdatedByAgentId: string | null;
+  latestRevisionId: string | null;
+  latestRevisionNumber: number;
+  revisionsToInsert: PlannedDocumentRevisionInsert[];
+  adjustments: ImportAdjustment[];
+};
+
+export type PlannedIssueDocumentSkip = {
+  source: IssueDocumentRow;
+  action: "skip_existing" | "skip_missing_parent" | "skip_conflicting_key";
+};
+
+export type PlannedAttachmentInsert = {
+  source: IssueAttachmentRow;
+  action: "insert";
+  targetIssueCommentId: string | null;
+  targetCreatedByAgentId: string | null;
+  adjustments: ImportAdjustment[];
+};
+
+export type PlannedAttachmentSkip = {
+  source: IssueAttachmentRow;
+  action: "skip_existing" | "skip_missing_parent";
+};
+
+export type WorktreeMergePlan = {
+  companyId: string;
+  companyName: string;
+  issuePrefix: string;
+  previewIssueCounterStart: number;
+  scopes: WorktreeMergeScope[];
+  issuePlans: Array<PlannedIssueInsert | PlannedIssueSkip>;
+  commentPlans: Array<PlannedCommentInsert | PlannedCommentSkip>;
+  documentPlans: Array<PlannedIssueDocumentInsert | PlannedIssueDocumentMerge | PlannedIssueDocumentSkip>;
+  attachmentPlans: Array<PlannedAttachmentInsert | PlannedAttachmentSkip>;
+  counts: {
+    issuesToInsert: number;
+    issuesExisting: number;
+    issueDrift: number;
+    commentsToInsert: number;
+    commentsExisting: number;
+    commentsMissingParent: number;
+    documentsToInsert: number;
+    documentsToMerge: number;
+    documentsExisting: number;
+    documentsConflictingKey: number;
+    documentsMissingParent: number;
+    documentRevisionsToInsert: number;
+    attachmentsToInsert: number;
+    attachmentsExisting: number;
+    attachmentsMissingParent: number;
+  };
+  adjustments: Record<ImportAdjustment, number>;
+};
+
+function compareIssueCoreFields(source: IssueRow, target: IssueRow): string[] {
+  const driftKeys: string[] = [];
+  if (source.title !== target.title) driftKeys.push("title");
+  if ((source.description ?? null) !== (target.description ?? null)) driftKeys.push("description");
+  if (source.status !== target.status) driftKeys.push("status");
+  if (source.priority !== target.priority) driftKeys.push("priority");
+  if ((source.parentId ?? null) !== (target.parentId ?? null)) driftKeys.push("parentId");
+  if ((source.projectId ?? null) !== (target.projectId ?? null)) driftKeys.push("projectId");
+  if ((source.projectWorkspaceId ?? null) !== (target.projectWorkspaceId ?? null)) driftKeys.push("projectWorkspaceId");
+  if ((source.goalId ?? null) !== (target.goalId ?? null)) driftKeys.push("goalId");
+  if ((source.assigneeAgentId ?? null) !== (target.assigneeAgentId ?? null)) driftKeys.push("assigneeAgentId");
+  if ((source.assigneeUserId ?? null) !== (target.assigneeUserId ?? null)) driftKeys.push("assigneeUserId");
+  return driftKeys;
+}
+
+function incrementAdjustment(
+  counts: Record<ImportAdjustment, number>,
+  adjustment: ImportAdjustment,
+): void {
+  counts[adjustment] += 1;
+}
+
+function groupBy<T>(rows: T[], keyFor: (row: T) => string): Map<string, T[]> {
+  const out = new Map<string, T[]>();
+  for (const row of rows) {
+    const key = keyFor(row);
+    const existing = out.get(key);
+    if (existing) {
+      existing.push(row);
+    } else {
+      out.set(key, [row]);
+    }
+  }
+  return out;
+}
+
+function sameDate(left: Date, right: Date): boolean {
+  return left.getTime() === right.getTime();
+}
+
+function sortDocumentRows(rows: IssueDocumentRow[]): IssueDocumentRow[] {
+  return [...rows].sort((left, right) => {
+    const createdDelta = left.documentCreatedAt.getTime() - right.documentCreatedAt.getTime();
+    if (createdDelta !== 0) return createdDelta;
+    const linkDelta = left.linkCreatedAt.getTime() - right.linkCreatedAt.getTime();
+    if (linkDelta !== 0) return linkDelta;
+    return left.documentId.localeCompare(right.documentId);
+  });
+}
+
+function sortDocumentRevisions(rows: DocumentRevisionRow[]): DocumentRevisionRow[] {
+  return [...rows].sort((left, right) => {
+    const revisionDelta = left.revisionNumber - right.revisionNumber;
+    if (revisionDelta !== 0) return revisionDelta;
+    const createdDelta = left.createdAt.getTime() - right.createdAt.getTime();
+    if (createdDelta !== 0) return createdDelta;
+    return left.id.localeCompare(right.id);
+  });
+}
+
+function sortAttachments(rows: IssueAttachmentRow[]): IssueAttachmentRow[] {
+  return [...rows].sort((left, right) => {
+    const createdDelta = left.attachmentCreatedAt.getTime() - right.attachmentCreatedAt.getTime();
+    if (createdDelta !== 0) return createdDelta;
+    return left.id.localeCompare(right.id);
+  });
+}
+
+function sortIssuesForImport(sourceIssues: IssueRow[]): IssueRow[] {
+  const byId = new Map(sourceIssues.map((issue) => [issue.id, issue]));
+  const memoDepth = new Map<string, number>();
+
+  const depthFor = (issue: IssueRow, stack = new Set<string>()): number => {
+    const memoized = memoDepth.get(issue.id);
+    if (memoized !== undefined) return memoized;
+    if (!issue.parentId) {
+      memoDepth.set(issue.id, 0);
+      return 0;
+    }
+    if (stack.has(issue.id)) {
+      memoDepth.set(issue.id, 0);
+      return 0;
+    }
+    const parent = byId.get(issue.parentId);
+    if (!parent) {
+      memoDepth.set(issue.id, 0);
+      return 0;
+    }
+    stack.add(issue.id);
+    const depth = depthFor(parent, stack) + 1;
+    stack.delete(issue.id);
+    memoDepth.set(issue.id, depth);
+    return depth;
+  };
+
+  return [...sourceIssues].sort((left, right) => {
+    const depthDelta = depthFor(left) - depthFor(right);
+    if (depthDelta !== 0) return depthDelta;
+    const createdDelta = left.createdAt.getTime() - right.createdAt.getTime();
+    if (createdDelta !== 0) return createdDelta;
+    return left.id.localeCompare(right.id);
+  });
+}
+
+export function parseWorktreeMergeScopes(rawValue: string | undefined): WorktreeMergeScope[] {
+  if (!rawValue || rawValue.trim().length === 0) {
+    return ["issues", "comments"];
+  }
+
+  const parsed = rawValue
+    .split(",")
+    .map((value) => value.trim().toLowerCase())
+    .filter((value): value is WorktreeMergeScope =>
+      (WORKTREE_MERGE_SCOPES as readonly string[]).includes(value),
+    );
+
+  if (parsed.length === 0) {
+    throw new Error(
+      `Invalid scope "${rawValue}". Expected a comma-separated list of: ${WORKTREE_MERGE_SCOPES.join(", ")}.`,
+    );
+  }
+
+  return [...new Set(parsed)];
+}
+
+export function buildWorktreeMergePlan(input: {
+  companyId: string;
+  companyName: string;
+  issuePrefix: string;
+  previewIssueCounterStart: number;
+  scopes: WorktreeMergeScope[];
+  sourceIssues: IssueRow[];
+  targetIssues: IssueRow[];
+  sourceComments: CommentRow[];
+  targetComments: CommentRow[];
+  sourceDocuments?: IssueDocumentRow[];
+  targetDocuments?: IssueDocumentRow[];
+  sourceDocumentRevisions?: DocumentRevisionRow[];
+  targetDocumentRevisions?: DocumentRevisionRow[];
+  sourceAttachments?: IssueAttachmentRow[];
+  targetAttachments?: IssueAttachmentRow[];
+  targetAgents: AgentRow[];
+  targetProjects: ProjectRow[];
+  targetProjectWorkspaces: ProjectWorkspaceRow[];
+  targetGoals: GoalRow[];
+  projectIdOverrides?: Record<string, string | null | undefined>;
+}): WorktreeMergePlan {
+  const targetIssuesById = new Map(input.targetIssues.map((issue) => [issue.id, issue]));
+  const targetCommentIds = new Set(input.targetComments.map((comment) => comment.id));
+  const targetAgentIds = new Set(input.targetAgents.map((agent) => agent.id));
+  const targetProjectIds = new Set(input.targetProjects.map((project) => project.id));
+  const targetProjectsById = new Map(input.targetProjects.map((project) => [project.id, project]));
+  const targetProjectWorkspaceIds = new Set(input.targetProjectWorkspaces.map((workspace) => workspace.id));
+  const targetGoalIds = new Set(input.targetGoals.map((goal) => goal.id));
+  const scopes = new Set(input.scopes);
+
+  const adjustmentCounts: Record<ImportAdjustment, number> = {
+    clear_assignee_agent: 0,
+    clear_project: 0,
+    clear_project_workspace: 0,
+    clear_goal: 0,
+    clear_author_agent: 0,
+    coerce_in_progress_to_todo: 0,
+    clear_document_agent: 0,
+    clear_document_revision_agent: 0,
+    clear_attachment_agent: 0,
+  };
+
+  const issuePlans: Array<PlannedIssueInsert | PlannedIssueSkip> = [];
+  let nextPreviewIssueNumber = input.previewIssueCounterStart;
+  for (const issue of sortIssuesForImport(input.sourceIssues)) {
+    const existing = targetIssuesById.get(issue.id);
+    if (existing) {
+      issuePlans.push({
+        source: issue,
+        action: "skip_existing",
+        driftKeys: compareIssueCoreFields(issue, existing),
+      });
+      continue;
+    }
+
+    nextPreviewIssueNumber += 1;
+    const adjustments: ImportAdjustment[] = [];
+    const targetAssigneeAgentId =
+      issue.assigneeAgentId && targetAgentIds.has(issue.assigneeAgentId) ? issue.assigneeAgentId : null;
+    if (issue.assigneeAgentId && !targetAssigneeAgentId) {
+      adjustments.push("clear_assignee_agent");
+      incrementAdjustment(adjustmentCounts, "clear_assignee_agent");
+    }
+
+    const targetCreatedByAgentId =
+      issue.createdByAgentId && targetAgentIds.has(issue.createdByAgentId) ? issue.createdByAgentId : null;
+
+    let targetProjectId =
+      issue.projectId && targetProjectIds.has(issue.projectId) ? issue.projectId : null;
+    let projectResolution: PlannedIssueInsert["projectResolution"] = targetProjectId ? "preserved" : "cleared";
+    let mappedProjectName: string | null = null;
+    const overrideProjectId =
+      issue.projectId && input.projectIdOverrides
+        ? input.projectIdOverrides[issue.projectId] ?? null
+        : null;
+    if (!targetProjectId && overrideProjectId && targetProjectIds.has(overrideProjectId)) {
+      targetProjectId = overrideProjectId;
+      projectResolution = "mapped";
+      mappedProjectName = targetProjectsById.get(overrideProjectId)?.name ?? null;
+    }
+    if (issue.projectId && !targetProjectId) {
+      adjustments.push("clear_project");
+      incrementAdjustment(adjustmentCounts, "clear_project");
+    }
+
+    const targetProjectWorkspaceId =
+      targetProjectId
+      && targetProjectId === issue.projectId
+      && issue.projectWorkspaceId
+      && targetProjectWorkspaceIds.has(issue.projectWorkspaceId)
+        ? issue.projectWorkspaceId
+        : null;
+    if (issue.projectWorkspaceId && !targetProjectWorkspaceId) {
+      adjustments.push("clear_project_workspace");
+      incrementAdjustment(adjustmentCounts, "clear_project_workspace");
+    }
+
+    const targetGoalId =
+      issue.goalId && targetGoalIds.has(issue.goalId) ? issue.goalId : null;
+    if (issue.goalId && !targetGoalId) {
+      adjustments.push("clear_goal");
+      incrementAdjustment(adjustmentCounts, "clear_goal");
+    }
+
+    let targetStatus = issue.status;
+    if (
+      targetStatus === "in_progress"
+      && !targetAssigneeAgentId
+      && !(issue.assigneeUserId && issue.assigneeUserId.trim().length > 0)
+    ) {
+      targetStatus = "todo";
+      adjustments.push("coerce_in_progress_to_todo");
+      incrementAdjustment(adjustmentCounts, "coerce_in_progress_to_todo");
+    }
+
+    issuePlans.push({
+      source: issue,
+      action: "insert",
+      previewIssueNumber: nextPreviewIssueNumber,
+      previewIdentifier: `${input.issuePrefix}-${nextPreviewIssueNumber}`,
+      targetStatus,
+      targetAssigneeAgentId,
+      targetCreatedByAgentId,
+      targetProjectId,
+      targetProjectWorkspaceId,
+      targetGoalId,
+      projectResolution,
+      mappedProjectName,
+      adjustments,
+    });
+  }
+
+  const issueIdsAvailableAfterImport = new Set<string>([
+    ...input.targetIssues.map((issue) => issue.id),
+    ...issuePlans.filter((plan): plan is PlannedIssueInsert => plan.action === "insert").map((plan) => plan.source.id),
+  ]);
+
+  const commentPlans: Array<PlannedCommentInsert | PlannedCommentSkip> = [];
+  if (scopes.has("comments")) {
+    const sortedComments = [...input.sourceComments].sort((left, right) => {
+      const createdDelta = left.createdAt.getTime() - right.createdAt.getTime();
+      if (createdDelta !== 0) return createdDelta;
+      return left.id.localeCompare(right.id);
+    });
+
+    for (const comment of sortedComments) {
+      if (targetCommentIds.has(comment.id)) {
+        commentPlans.push({ source: comment, action: "skip_existing" });
+        continue;
+      }
+      if (!issueIdsAvailableAfterImport.has(comment.issueId)) {
+        commentPlans.push({ source: comment, action: "skip_missing_parent" });
+        continue;
+      }
+
+      const adjustments: ImportAdjustment[] = [];
+      const targetAuthorAgentId =
+        comment.authorAgentId && targetAgentIds.has(comment.authorAgentId) ? comment.authorAgentId : null;
+      if (comment.authorAgentId && !targetAuthorAgentId) {
+        adjustments.push("clear_author_agent");
+        incrementAdjustment(adjustmentCounts, "clear_author_agent");
+      }
+
+      commentPlans.push({
+        source: comment,
+        action: "insert",
+        targetAuthorAgentId,
+        adjustments,
+      });
+    }
+  }
+
+  const sourceDocuments = input.sourceDocuments ?? [];
+  const targetDocuments = input.targetDocuments ?? [];
+  const sourceDocumentRevisions = input.sourceDocumentRevisions ?? [];
+  const targetDocumentRevisions = input.targetDocumentRevisions ?? [];
+
+  const targetDocumentsById = new Map(targetDocuments.map((document) => [document.documentId, document]));
+  const targetDocumentsByIssueKey = new Map(targetDocuments.map((document) => [`${document.issueId}:${document.key}`, document]));
+  const sourceRevisionsByDocumentId = groupBy(sourceDocumentRevisions, (revision) => revision.documentId);
+  const targetRevisionsByDocumentId = groupBy(targetDocumentRevisions, (revision) => revision.documentId);
+  const commentIdsAvailableAfterImport = new Set<string>([
+    ...input.targetComments.map((comment) => comment.id),
+    ...commentPlans.filter((plan): plan is PlannedCommentInsert => plan.action === "insert").map((plan) => plan.source.id),
+  ]);
+
+  const documentPlans: Array<PlannedIssueDocumentInsert | PlannedIssueDocumentMerge | PlannedIssueDocumentSkip> = [];
+  for (const document of sortDocumentRows(sourceDocuments)) {
+    if (!issueIdsAvailableAfterImport.has(document.issueId)) {
+      documentPlans.push({ source: document, action: "skip_missing_parent" });
+      continue;
+    }
+
+    const existingDocument = targetDocumentsById.get(document.documentId);
+    const conflictingIssueKeyDocument = targetDocumentsByIssueKey.get(`${document.issueId}:${document.key}`);
+    if (!existingDocument && conflictingIssueKeyDocument && conflictingIssueKeyDocument.documentId !== document.documentId) {
+      documentPlans.push({ source: document, action: "skip_conflicting_key" });
+      continue;
+    }
+
+    const adjustments: ImportAdjustment[] = [];
+    const targetCreatedByAgentId =
+      document.createdByAgentId && targetAgentIds.has(document.createdByAgentId) ? document.createdByAgentId : null;
+    const targetUpdatedByAgentId =
+      document.updatedByAgentId && targetAgentIds.has(document.updatedByAgentId) ? document.updatedByAgentId : null;
+    if (
+      (document.createdByAgentId && !targetCreatedByAgentId)
+      || (document.updatedByAgentId && !targetUpdatedByAgentId)
+    ) {
+      adjustments.push("clear_document_agent");
+      incrementAdjustment(adjustmentCounts, "clear_document_agent");
+    }
+
+    const sourceRevisions = sortDocumentRevisions(sourceRevisionsByDocumentId.get(document.documentId) ?? []);
+    const targetRevisions = sortDocumentRevisions(targetRevisionsByDocumentId.get(document.documentId) ?? []);
+    const existingRevisionIds = new Set(targetRevisions.map((revision) => revision.id));
+    const usedRevisionNumbers = new Set(targetRevisions.map((revision) => revision.revisionNumber));
+    let nextRevisionNumber = targetRevisions.reduce(
+      (maxValue, revision) => Math.max(maxValue, revision.revisionNumber),
+      0,
+    ) + 1;
+
+    const targetRevisionNumberById = new Map<string, number>(
+      targetRevisions.map((revision) => [revision.id, revision.revisionNumber]),
+    );
+    const revisionsToInsert: PlannedDocumentRevisionInsert[] = [];
+
+    for (const revision of sourceRevisions) {
+      if (existingRevisionIds.has(revision.id)) continue;
+      let targetRevisionNumber = revision.revisionNumber;
+      if (usedRevisionNumbers.has(targetRevisionNumber)) {
+        while (usedRevisionNumbers.has(nextRevisionNumber)) {
+          nextRevisionNumber += 1;
+        }
+        targetRevisionNumber = nextRevisionNumber;
+        nextRevisionNumber += 1;
+      }
+      usedRevisionNumbers.add(targetRevisionNumber);
+      targetRevisionNumberById.set(revision.id, targetRevisionNumber);
+
+      const revisionAdjustments: ImportAdjustment[] = [];
+      const targetCreatedByAgentId =
+        revision.createdByAgentId && targetAgentIds.has(revision.createdByAgentId) ? revision.createdByAgentId : null;
+      if (revision.createdByAgentId && !targetCreatedByAgentId) {
+        revisionAdjustments.push("clear_document_revision_agent");
+        incrementAdjustment(adjustmentCounts, "clear_document_revision_agent");
+      }
+
+      revisionsToInsert.push({
+        source: revision,
+        targetRevisionNumber,
+        targetCreatedByAgentId,
+        adjustments: revisionAdjustments,
+      });
+    }
+
+    const latestRevisionId = document.latestRevisionId ?? existingDocument?.latestRevisionId ?? null;
+    const latestRevisionNumber =
+      (latestRevisionId ? targetRevisionNumberById.get(latestRevisionId) : undefined)
+      ?? document.latestRevisionNumber
+      ?? existingDocument?.latestRevisionNumber
+      ?? 0;
+
+    if (!existingDocument) {
+      documentPlans.push({
+        source: document,
+        action: "insert",
+        targetCreatedByAgentId,
+        targetUpdatedByAgentId,
+        latestRevisionId,
+        latestRevisionNumber,
+        revisionsToInsert,
+        adjustments,
+      });
+      continue;
+    }
+
+    const documentAlreadyMatches =
+      existingDocument.key === document.key
+      && existingDocument.title === document.title
+      && existingDocument.format === document.format
+      && existingDocument.latestBody === document.latestBody
+      && (existingDocument.latestRevisionId ?? null) === latestRevisionId
+      && existingDocument.latestRevisionNumber === latestRevisionNumber
+      && (existingDocument.updatedByAgentId ?? null) === targetUpdatedByAgentId
+      && (existingDocument.updatedByUserId ?? null) === (document.updatedByUserId ?? null)
+      && sameDate(existingDocument.documentUpdatedAt, document.documentUpdatedAt)
+      && sameDate(existingDocument.linkUpdatedAt, document.linkUpdatedAt)
+      && revisionsToInsert.length === 0;
+
+    if (documentAlreadyMatches) {
+      documentPlans.push({ source: document, action: "skip_existing" });
+      continue;
+    }
+
+    documentPlans.push({
+      source: document,
+      action: "merge_existing",
+      targetCreatedByAgentId,
+      targetUpdatedByAgentId,
+      latestRevisionId,
+      latestRevisionNumber,
+      revisionsToInsert,
+      adjustments,
+    });
+  }
+
+  const sourceAttachments = input.sourceAttachments ?? [];
+  const targetAttachmentIds = new Set((input.targetAttachments ?? []).map((attachment) => attachment.id));
+  const attachmentPlans: Array<PlannedAttachmentInsert | PlannedAttachmentSkip> = [];
+  for (const attachment of sortAttachments(sourceAttachments)) {
+    if (targetAttachmentIds.has(attachment.id)) {
+      attachmentPlans.push({ source: attachment, action: "skip_existing" });
+      continue;
+    }
+    if (!issueIdsAvailableAfterImport.has(attachment.issueId)) {
+      attachmentPlans.push({ source: attachment, action: "skip_missing_parent" });
+      continue;
+    }
+
+    const adjustments: ImportAdjustment[] = [];
+    const targetCreatedByAgentId =
+      attachment.createdByAgentId && targetAgentIds.has(attachment.createdByAgentId)
+        ? attachment.createdByAgentId
+        : null;
+    if (attachment.createdByAgentId && !targetCreatedByAgentId) {
+      adjustments.push("clear_attachment_agent");
+      incrementAdjustment(adjustmentCounts, "clear_attachment_agent");
+    }
+
+    attachmentPlans.push({
+      source: attachment,
+      action: "insert",
+      targetIssueCommentId:
+        attachment.issueCommentId && commentIdsAvailableAfterImport.has(attachment.issueCommentId)
+          ? attachment.issueCommentId
+          : null,
+      targetCreatedByAgentId,
+      adjustments,
+    });
+  }
+
+  const counts = {
+    issuesToInsert: issuePlans.filter((plan) => plan.action === "insert").length,
+    issuesExisting: issuePlans.filter((plan) => plan.action === "skip_existing").length,
+    issueDrift: issuePlans.filter((plan) => plan.action === "skip_existing" && plan.driftKeys.length > 0).length,
+    commentsToInsert: commentPlans.filter((plan) => plan.action === "insert").length,
+    commentsExisting: commentPlans.filter((plan) => plan.action === "skip_existing").length,
+    commentsMissingParent: commentPlans.filter((plan) => plan.action === "skip_missing_parent").length,
+    documentsToInsert: documentPlans.filter((plan) => plan.action === "insert").length,
+    documentsToMerge: documentPlans.filter((plan) => plan.action === "merge_existing").length,
+    documentsExisting: documentPlans.filter((plan) => plan.action === "skip_existing").length,
+    documentsConflictingKey: documentPlans.filter((plan) => plan.action === "skip_conflicting_key").length,
+    documentsMissingParent: documentPlans.filter((plan) => plan.action === "skip_missing_parent").length,
+    documentRevisionsToInsert: documentPlans.reduce(
+      (sum, plan) =>
+        sum + (plan.action === "insert" || plan.action === "merge_existing" ? plan.revisionsToInsert.length : 0),
+      0,
+    ),
+    attachmentsToInsert: attachmentPlans.filter((plan) => plan.action === "insert").length,
+    attachmentsExisting: attachmentPlans.filter((plan) => plan.action === "skip_existing").length,
+    attachmentsMissingParent: attachmentPlans.filter((plan) => plan.action === "skip_missing_parent").length,
+  };
+
+  return {
+    companyId: input.companyId,
+    companyName: input.companyName,
+    issuePrefix: input.issuePrefix,
+    previewIssueCounterStart: input.previewIssueCounterStart,
+    scopes: input.scopes,
+    issuePlans,
+    commentPlans,
+    documentPlans,
+    attachmentPlans,
+    counts,
+    adjustments: adjustmentCounts,
+  };
+}

cli/src/config/env.ts

@@ -22,20 +22,35 @@ function parseEnvFile(contents: string) {
   }
 }
 
+function formatEnvValue(value: string): string {
+  if (/^[A-Za-z0-9_./:@-]+$/.test(value)) {
+    return value;
+  }
+  return JSON.stringify(value);
+}
+
 function renderEnvFile(entries: Record<string, string>) {
   const lines = [
     "# Paperclip environment variables",
-    "# Generated by `paperclipai onboard`",
-    ...Object.entries(entries).map(([key, value]) => `${key}=${value}`),
+    "# Generated by Paperclip CLI commands",
+    ...Object.entries(entries).map(([key, value]) => `${key}=${formatEnvValue(value)}`),
     "",
   ];
   return lines.join("\n");
 }
 
+export function resolvePaperclipEnvFile(configPath?: string): string {
+  return resolveEnvFilePath(configPath);
+}
+
 export function resolveAgentJwtEnvFile(configPath?: string): string {
   return resolveEnvFilePath(configPath);
 }
 
+export function loadPaperclipEnvFile(configPath?: string): void {
+  loadAgentJwtEnvFile(resolveEnvFilePath(configPath));
+}
+
 export function loadAgentJwtEnvFile(filePath = resolveEnvFilePath()): void {
   if (loadedEnvFiles.has(filePath)) return;
 
@@ -78,13 +93,33 @@ export function ensureAgentJwtSecret(configPath?: string): { secret: string; cre
 }
 
 export function writeAgentJwtEnv(secret: string, filePath = resolveEnvFilePath()): void {
+  mergePaperclipEnvEntries({ [JWT_SECRET_ENV_KEY]: secret }, filePath);
+}
+
+export function readPaperclipEnvEntries(filePath = resolveEnvFilePath()): Record<string, string> {
+  if (!fs.existsSync(filePath)) return {};
+  return parseEnvFile(fs.readFileSync(filePath, "utf-8"));
+}
+
+export function writePaperclipEnvEntries(entries: Record<string, string>, filePath = resolveEnvFilePath()): void {
   const dir = path.dirname(filePath);
   fs.mkdirSync(dir, { recursive: true });
-
-  const current = fs.existsSync(filePath) ? parseEnvFile(fs.readFileSync(filePath, "utf-8")) : {};
-  current[JWT_SECRET_ENV_KEY] = secret;
-
-  fs.writeFileSync(filePath, renderEnvFile(current), {
+  fs.writeFileSync(filePath, renderEnvFile(entries), {
     mode: 0o600,
   });
 }
+
+export function mergePaperclipEnvEntries(
+  entries: Record<string, string>,
+  filePath = resolveEnvFilePath(),
+): Record<string, string> {
+  const current = readPaperclipEnvEntries(filePath);
+  const next = {
+    ...current,
+    ...Object.fromEntries(
+      Object.entries(entries).filter(([, value]) => typeof value === "string" && value.trim().length > 0),
+    ),
+  };
+  writePaperclipEnvEntries(next, filePath);
+  return next;
+}

cli/src/config/home.ts

@@ -49,6 +49,10 @@ export function resolveDefaultStorageDir(instanceId?: string): string {
   return path.resolve(resolvePaperclipInstanceRoot(instanceId), "data", "storage");
 }
 
+export function resolveDefaultBackupDir(instanceId?: string): string {
+  return path.resolve(resolvePaperclipInstanceRoot(instanceId), "data", "backups");
+}
+
 export function expandHomePrefix(value: string): string {
   if (value === "~") return os.homedir();
   if (value.startsWith("~/")) return path.resolve(os.homedir(), value.slice(2));
@@ -64,6 +68,7 @@ export function describeLocalInstancePaths(instanceId?: string) {
     instanceRoot,
     configPath: resolveDefaultConfigPath(resolvedInstanceId),
     embeddedPostgresDataDir: resolveDefaultEmbeddedPostgresDir(resolvedInstanceId),
+    backupDir: resolveDefaultBackupDir(resolvedInstanceId),
     logDir: resolveDefaultLogsDir(resolvedInstanceId),
     secretsKeyFilePath: resolveDefaultSecretsKeyFilePath(resolvedInstanceId),
     storageDir: resolveDefaultStorageDir(resolvedInstanceId),

cli/src/config/schema.ts

@@ -2,6 +2,7 @@ export {
   paperclipConfigSchema,
   configMetaSchema,
   llmConfigSchema,
+  databaseBackupConfigSchema,
   databaseConfigSchema,
   loggingConfigSchema,
   serverConfigSchema,
@@ -13,6 +14,7 @@ export {
   secretsLocalEncryptedConfigSchema,
   type PaperclipConfig,
   type LlmConfig,
+  type DatabaseBackupConfig,
   type DatabaseConfig,
   type LoggingConfig,
   type ServerConfig,

cli/src/index.ts

@@ -7,6 +7,7 @@ import { addAllowedHostname } from "./commands/allowed-hostname.js";
 import { heartbeatRun } from "./commands/heartbeat-run.js";
 import { runCommand } from "./commands/run.js";
 import { bootstrapCeoInvite } from "./commands/auth-bootstrap-ceo.js";
+import { dbBackupCommand } from "./commands/db-backup.js";
 import { registerContextCommands } from "./commands/client/context.js";
 import { registerCompanyCommands } from "./commands/client/company.js";
 import { registerIssueCommands } from "./commands/client/issue.js";
@@ -15,6 +16,9 @@ import { registerApprovalCommands } from "./commands/client/approval.js";
 import { registerActivityCommands } from "./commands/client/activity.js";
 import { registerDashboardCommands } from "./commands/client/dashboard.js";
 import { applyDataDirOverride, type DataDirOptionLike } from "./config/data-dir.js";
+import { loadPaperclipEnvFile } from "./config/env.js";
+import { registerWorktreeCommands } from "./commands/worktree.js";
+import { registerPluginCommands } from "./commands/client/plugin.js";
 
 const program = new Command();
 const DATA_DIR_OPTION_HELP =
@@ -23,7 +27,7 @@ const DATA_DIR_OPTION_HELP =
 program
   .name("paperclipai")
   .description("Paperclip CLI — setup, diagnose, and configure your instance")
-  .version("0.2.5");
+  .version("0.2.7");
 
 program.hook("preAction", (_thisCommand, actionCommand) => {
   const options = actionCommand.optsWithGlobals() as DataDirOptionLike;
@@ -32,6 +36,7 @@ program.hook("preAction", (_thisCommand, actionCommand) => {
     hasConfigOption: optionNames.has("config"),
     hasContextOption: optionNames.has("context"),
   });
+  loadPaperclipEnvFile(options.config);
 });
 
 program
@@ -70,6 +75,19 @@ program
   .option("-s, --section <section>", "Section to configure (llm, database, logging, server, storage, secrets)")
   .action(configure);
 
+program
+  .command("db:backup")
+  .description("Create a one-off database backup using current config")
+  .option("-c, --config <path>", "Path to config file")
+  .option("-d, --data-dir <path>", DATA_DIR_OPTION_HELP)
+  .option("--dir <path>", "Backup output directory (overrides config)")
+  .option("--retention-days <days>", "Retention window used for pruning", (value) => Number(value))
+  .option("--filename-prefix <prefix>", "Backup filename prefix", "paperclip")
+  .option("--json", "Print backup metadata as JSON")
+  .action(async (opts) => {
+    await dbBackupCommand(opts);
+  });
+
 program
   .command("allowed-hostname")
   .description("Allow a hostname for authenticated/private mode access")
@@ -118,6 +136,8 @@ registerAgentCommands(program);
 registerApprovalCommands(program);
 registerActivityCommands(program);
 registerDashboardCommands(program);
+registerWorktreeCommands(program);
+registerPluginCommands(program);
 
 const auth = program.command("auth").description("Authentication and bootstrap utilities");
 

cli/src/prompts/database.ts

@@ -1,9 +1,26 @@
 import * as p from "@clack/prompts";
 import type { DatabaseConfig } from "../config/schema.js";
-import { resolveDefaultEmbeddedPostgresDir, resolvePaperclipInstanceId } from "../config/home.js";
+import {
+  resolveDefaultBackupDir,
+  resolveDefaultEmbeddedPostgresDir,
+  resolvePaperclipInstanceId,
+} from "../config/home.js";
 
-export async function promptDatabase(): Promise<DatabaseConfig> {
-  const defaultEmbeddedDir = resolveDefaultEmbeddedPostgresDir(resolvePaperclipInstanceId());
+export async function promptDatabase(current?: DatabaseConfig): Promise<DatabaseConfig> {
+  const instanceId = resolvePaperclipInstanceId();
+  const defaultEmbeddedDir = resolveDefaultEmbeddedPostgresDir(instanceId);
+  const defaultBackupDir = resolveDefaultBackupDir(instanceId);
+  const base: DatabaseConfig = current ?? {
+    mode: "embedded-postgres",
+    embeddedPostgresDataDir: defaultEmbeddedDir,
+    embeddedPostgresPort: 54329,
+    backup: {
+      enabled: true,
+      intervalMinutes: 60,
+      retentionDays: 30,
+      dir: defaultBackupDir,
+    },
+  };
 
   const mode = await p.select({
     message: "Database mode",
@@ -11,6 +28,7 @@ export async function promptDatabase(): Promise<DatabaseConfig> {
       { value: "embedded-postgres" as const, label: "Embedded PostgreSQL (managed locally)", hint: "recommended" },
       { value: "postgres" as const, label: "PostgreSQL (external server)" },
     ],
+    initialValue: base.mode,
   });
 
   if (p.isCancel(mode)) {
@@ -18,9 +36,14 @@ export async function promptDatabase(): Promise<DatabaseConfig> {
     process.exit(0);
   }
 
+  let connectionString: string | undefined = base.connectionString;
+  let embeddedPostgresDataDir = base.embeddedPostgresDataDir || defaultEmbeddedDir;
+  let embeddedPostgresPort = base.embeddedPostgresPort || 54329;
+
   if (mode === "postgres") {
-    const connectionString = await p.text({
+    const value = await p.text({
       message: "PostgreSQL connection string",
+      defaultValue: base.connectionString ?? "",
       placeholder: "postgres://user:pass@localhost:5432/paperclip",
       validate: (val) => {
         if (!val) return "Connection string is required for PostgreSQL mode";
@@ -28,48 +51,107 @@ export async function promptDatabase(): Promise<DatabaseConfig> {
       },
     });
 
-    if (p.isCancel(connectionString)) {
+    if (p.isCancel(value)) {
       p.cancel("Setup cancelled.");
       process.exit(0);
     }
 
-    return {
-      mode: "postgres",
-      connectionString,
-      embeddedPostgresDataDir: defaultEmbeddedDir,
-      embeddedPostgresPort: 54329,
-    };
+    connectionString = value;
+  } else {
+    const dataDir = await p.text({
+      message: "Embedded PostgreSQL data directory",
+      defaultValue: base.embeddedPostgresDataDir || defaultEmbeddedDir,
+      placeholder: defaultEmbeddedDir,
+    });
+
+    if (p.isCancel(dataDir)) {
+      p.cancel("Setup cancelled.");
+      process.exit(0);
+    }
+
+    embeddedPostgresDataDir = dataDir || defaultEmbeddedDir;
+
+    const portValue = await p.text({
+      message: "Embedded PostgreSQL port",
+      defaultValue: String(base.embeddedPostgresPort || 54329),
+      placeholder: "54329",
+      validate: (val) => {
+        const n = Number(val);
+        if (!Number.isInteger(n) || n < 1 || n > 65535) return "Port must be an integer between 1 and 65535";
+      },
+    });
+
+    if (p.isCancel(portValue)) {
+      p.cancel("Setup cancelled.");
+      process.exit(0);
+    }
+
+    embeddedPostgresPort = Number(portValue || "54329");
+    connectionString = undefined;
   }
 
-  const embeddedPostgresDataDir = await p.text({
-    message: "Embedded PostgreSQL data directory",
-    defaultValue: defaultEmbeddedDir,
-    placeholder: defaultEmbeddedDir,
+  const backupEnabled = await p.confirm({
+    message: "Enable automatic database backups?",
+    initialValue: base.backup.enabled,
   });
-
-  if (p.isCancel(embeddedPostgresDataDir)) {
+  if (p.isCancel(backupEnabled)) {
     p.cancel("Setup cancelled.");
     process.exit(0);
   }
 
-  const embeddedPostgresPort = await p.text({
-    message: "Embedded PostgreSQL port",
-    defaultValue: "54329",
-    placeholder: "54329",
+  const backupDirInput = await p.text({
+    message: "Backup directory",
+    defaultValue: base.backup.dir || defaultBackupDir,
+    placeholder: defaultBackupDir,
+    validate: (val) => (!val || val.trim().length === 0 ? "Backup directory is required" : undefined),
+  });
+  if (p.isCancel(backupDirInput)) {
+    p.cancel("Setup cancelled.");
+    process.exit(0);
+  }
+
+  const backupIntervalInput = await p.text({
+    message: "Backup interval (minutes)",
+    defaultValue: String(base.backup.intervalMinutes || 60),
+    placeholder: "60",
     validate: (val) => {
       const n = Number(val);
-      if (!Number.isInteger(n) || n < 1 || n > 65535) return "Port must be an integer between 1 and 65535";
+      if (!Number.isInteger(n) || n < 1) return "Interval must be a positive integer";
+      if (n > 10080) return "Interval must be 10080 minutes (7 days) or less";
+      return undefined;
     },
   });
+  if (p.isCancel(backupIntervalInput)) {
+    p.cancel("Setup cancelled.");
+    process.exit(0);
+  }
 
-  if (p.isCancel(embeddedPostgresPort)) {
+  const backupRetentionInput = await p.text({
+    message: "Backup retention (days)",
+    defaultValue: String(base.backup.retentionDays || 30),
+    placeholder: "30",
+    validate: (val) => {
+      const n = Number(val);
+      if (!Number.isInteger(n) || n < 1) return "Retention must be a positive integer";
+      if (n > 3650) return "Retention must be 3650 days or less";
+      return undefined;
+    },
+  });
+  if (p.isCancel(backupRetentionInput)) {
     p.cancel("Setup cancelled.");
     process.exit(0);
   }
 
   return {
-    mode: "embedded-postgres",
-    embeddedPostgresDataDir: embeddedPostgresDataDir || defaultEmbeddedDir,
-    embeddedPostgresPort: Number(embeddedPostgresPort || "54329"),
+    mode,
+    connectionString,
+    embeddedPostgresDataDir,
+    embeddedPostgresPort,
+    backup: {
+      enabled: backupEnabled,
+      intervalMinutes: Number(backupIntervalInput || "60"),
+      retentionDays: Number(backupRetentionInput || "30"),
+      dir: backupDirInput || defaultBackupDir,
+    },
   };
 }

cli/src/prompts/server.ts

@@ -113,7 +113,7 @@ export async function promptServer(opts?: {
   }
 
   const port = Number(portStr) || 3100;
-  let auth: AuthConfig = { baseUrlMode: "auto" };
+  let auth: AuthConfig = { baseUrlMode: "auto", disableSignUp: false };
   if (deploymentMode === "authenticated" && exposure === "public") {
     const urlInput = await p.text({
       message: "Public base URL",
@@ -139,18 +139,26 @@ export async function promptServer(opts?: {
     }
     auth = {
       baseUrlMode: "explicit",
+      disableSignUp: false,
       publicBaseUrl: urlInput.trim().replace(/\/+$/, ""),
     };
   } else if (currentAuth?.baseUrlMode === "explicit" && currentAuth.publicBaseUrl) {
     auth = {
       baseUrlMode: "explicit",
+      disableSignUp: false,
       publicBaseUrl: currentAuth.publicBaseUrl,
     };
   }
 
   return {
-    server: { deploymentMode, exposure, host: hostStr.trim(), port, allowedHostnames, serveUi: true },
+    server: {
+      deploymentMode,
+      exposure,
+      host: hostStr.trim(),
+      port,
+      allowedHostnames,
+      serveUi: currentServer?.serveUi ?? true,
+    },
     auth,
   };
 }
-

cli/tsconfig.json

@@ -1,5 +1,5 @@
 {
-  "extends": "../tsconfig.json",
+  "extends": "../tsconfig.base.json",
   "compilerOptions": {
     "outDir": "dist",
     "rootDir": "src"

doc/AGENTCOMPANIES_SPEC_INVENTORY.md

@@ -0,0 +1,114 @@
+# Agent Companies Spec Inventory
+
+This document indexes every part of the Paperclip codebase that touches the [Agent Companies Specification](docs/companies/companies-spec.md) (`agentcompanies/v1-draft`).
+
+Use it when you need to:
+
+1. **Update the spec** — know which implementation code must change in lockstep.
+2. **Change code that involves the spec** — find all related files quickly.
+3. **Keep things aligned** — audit whether implementation matches the spec.
+
+---
+
+## 1. Specification & Design Documents
+
+| File | Role |
+|---|---|
+| `docs/companies/companies-spec.md` | **Normative spec** — defines the markdown-first package format (COMPANY.md, TEAM.md, AGENTS.md, PROJECT.md, TASK.md, SKILL.md), reserved files, frontmatter schemas, and vendor extension conventions (`.paperclip.yaml`). |
+| `doc/plans/2026-03-13-company-import-export-v2.md` | Implementation plan for the markdown-first package model cutover — phases, API changes, UI plan, and rollout strategy. |
+| `doc/SPEC-implementation.md` | V1 implementation contract; references the portability system and `.paperclip.yaml` sidecar format. |
+| `docs/specs/cliphub-plan.md` | Earlier blueprint bundle plan; partially superseded by the markdown-first spec (noted in the v2 plan). |
+| `doc/plans/2026-02-16-module-system.md` | Module system plan; JSON-only company template sections superseded by the markdown-first model. |
+| `doc/plans/2026-03-14-skills-ui-product-plan.md` | Skills UI plan; references portable skill files and `.paperclip.yaml`. |
+| `doc/plans/2026-03-14-adapter-skill-sync-rollout.md` | Adapter skill sync rollout; companion to the v2 import/export plan. |
+
+## 2. Shared Types & Validators
+
+These define the contract between server, CLI, and UI.
+
+| File | What it defines |
+|---|---|
+| `packages/shared/src/types/company-portability.ts` | TypeScript interfaces: `CompanyPortabilityManifest`, `CompanyPortabilityFileEntry`, `CompanyPortabilityEnvInput`, export/import/preview request and result types, manifest entry types for agents, skills, projects, issues, companies. |
+| `packages/shared/src/validators/company-portability.ts` | Zod schemas for all portability request/response shapes — used by both server routes and CLI. |
+| `packages/shared/src/types/index.ts` | Re-exports portability types. |
+| `packages/shared/src/validators/index.ts` | Re-exports portability validators. |
+
+## 3. Server — Services
+
+| File | Responsibility |
+|---|---|
+| `server/src/services/company-portability.ts` | **Core portability service.** Export (manifest generation, markdown file emission, `.paperclip.yaml` sidecars), import (graph resolution, collision handling, entity creation), preview (planned-action summary). Handles skill key derivation, task recurrence parsing, and package README generation. References `agentcompanies/v1` version string. |
+| `server/src/services/company-export-readme.ts` | Generates `README.md` and Mermaid org-chart for exported company packages. |
+| `server/src/services/index.ts` | Re-exports `companyPortabilityService`. |
+
+## 4. Server — Routes
+
+| File | Endpoints |
+|---|---|
+| `server/src/routes/companies.ts` | `POST /api/companies/:companyId/export` — legacy export bundle<br>`POST /api/companies/:companyId/exports/preview` — export preview<br>`POST /api/companies/:companyId/exports` — export package<br>`POST /api/companies/import/preview` — import preview<br>`POST /api/companies/import` — perform import |
+
+Route registration lives in `server/src/app.ts` via `companyRoutes(db, storage)`.
+
+## 5. Server — Tests
+
+| File | Coverage |
+|---|---|
+| `server/src/__tests__/company-portability.test.ts` | Unit tests for the portability service (export, import, preview, manifest shape, `agentcompanies/v1` version). |
+| `server/src/__tests__/company-portability-routes.test.ts` | Integration tests for the portability HTTP endpoints. |
+
+## 6. CLI
+
+| File | Commands |
+|---|---|
+| `cli/src/commands/client/company.ts` | `company export` — exports a company package to disk (flags: `--out`, `--include`, `--projects`, `--issues`, `--projectIssues`).<br>`company import` — imports a company package from a file or folder (flags: `--from`, `--include`, `--target`, `--companyId`, `--newCompanyName`, `--agents`, `--collision`, `--dryRun`).<br>Reads/writes portable file entries and handles `.paperclip.yaml` filtering. |
+
+## 7. UI — Pages
+
+| File | Role |
+|---|---|
+| `ui/src/pages/CompanyExport.tsx` | Export UI: preview, manifest display, file tree visualization, ZIP archive creation and download. Filters `.paperclip.yaml` based on selection. Shows manifest and README in editor. |
+| `ui/src/pages/CompanyImport.tsx` | Import UI: source input (upload/folder/GitHub URL/generic URL), ZIP reading, preview pane with dependency tree, entity selection checkboxes, trust/licensing warnings, secrets requirements, collision strategy, adapter config. |
+
+## 8. UI — Components
+
+| File | Role |
+|---|---|
+| `ui/src/components/PackageFileTree.tsx` | Reusable file tree component for both import and export. Builds tree from `CompanyPortabilityFileEntry` items, parses frontmatter, shows action indicators (create/update/skip), and maps frontmatter field labels. |
+
+## 9. UI — Libraries
+
+| File | Role |
+|---|---|
+| `ui/src/lib/portable-files.ts` | Helpers for portable file entries: `getPortableFileText`, `getPortableFileDataUrl`, `getPortableFileContentType`, `isPortableImageFile`. |
+| `ui/src/lib/zip.ts` | ZIP archive creation (`createZipArchive`) and reading (`readZipArchive`) — implements ZIP format from scratch for company packages. CRC32, DOS date/time encoding. |
+| `ui/src/lib/zip.test.ts` | Tests for ZIP utilities; exercises round-trip with portability file entries and `.paperclip.yaml` content. |
+
+## 10. UI — API Client
+
+| File | Functions |
+|---|---|
+| `ui/src/api/companies.ts` | `companiesApi.exportBundle`, `companiesApi.exportPreview`, `companiesApi.exportPackage`, `companiesApi.importPreview`, `companiesApi.importBundle` — typed fetch wrappers for the portability endpoints. |
+
+## 11. Skills & Agent Instructions
+
+| File | Relevance |
+|---|---|
+| `skills/paperclip/references/company-skills.md` | Reference doc for company skill library workflow — install, inspect, update, assign. Skill packages are a subset of the agent companies spec. |
+| `server/src/services/company-skills.ts` | Company skill management service — handles SKILL.md-based imports and company-level skill library. |
+| `server/src/services/agent-instructions.ts` | Agent instructions service — resolves AGENTS.md paths for agent instruction loading. |
+
+## 12. Quick Cross-Reference by Spec Concept
+
+| Spec concept | Primary implementation files |
+|---|---|
+| `COMPANY.md` frontmatter & body | `company-portability.ts` (export emitter + import parser) |
+| `AGENTS.md` frontmatter & body | `company-portability.ts`, `agent-instructions.ts` |
+| `PROJECT.md` frontmatter & body | `company-portability.ts` |
+| `TASK.md` frontmatter & body | `company-portability.ts` |
+| `SKILL.md` packages | `company-portability.ts`, `company-skills.ts` |
+| `.paperclip.yaml` vendor sidecar | `company-portability.ts`, `CompanyExport.tsx`, `company.ts` (CLI) |
+| `manifest.json` | `company-portability.ts` (generation), shared types (schema) |
+| ZIP package format | `zip.ts` (UI), `company.ts` (CLI file I/O) |
+| Collision resolution | `company-portability.ts` (server), `CompanyImport.tsx` (UI) |
+| Env/secrets declarations | shared types (`CompanyPortabilityEnvInput`), `CompanyImport.tsx` (UI) |
+| README + org chart | `company-export-readme.ts` |

doc/CLI.md

@@ -116,6 +116,20 @@ pnpm paperclipai issue release <issue-id>
 ```sh
 pnpm paperclipai agent list --company-id <company-id>
 pnpm paperclipai agent get <agent-id>
+pnpm paperclipai agent local-cli <agent-id-or-shortname> --company-id <company-id>
+```
+
+`agent local-cli` is the quickest way to run local Claude/Codex manually as a Paperclip agent:
+
+- creates a new long-lived agent API key
+- installs missing Paperclip skills into `~/.codex/skills` and `~/.claude/skills`
+- prints `export ...` lines for `PAPERCLIP_API_URL`, `PAPERCLIP_COMPANY_ID`, `PAPERCLIP_AGENT_ID`, and `PAPERCLIP_API_KEY`
+
+Example for shortname-based local setup:
+
+```sh
+pnpm paperclipai agent local-cli codexcoder --company-id <company-id>
+pnpm paperclipai agent local-cli claudecoder --company-id <company-id>
 ```
 
 ## Approval Commands

doc/DATABASE.md

@@ -19,6 +19,14 @@ That's it. On first start the server:
 
 Data persists across restarts in `~/.paperclip/instances/default/db/`. To reset local dev data, delete that directory.
 
+If you need to apply pending migrations manually, run:
+
+```sh
+pnpm db:migrate
+```
+
+When `DATABASE_URL` is unset, this command targets the current embedded PostgreSQL instance for your active Paperclip config/instance.
+
 This mode is ideal for local development and one-command installs.
 
 Docker note: the Docker quickstart image also uses embedded PostgreSQL by default. Persist `/paperclip` to keep DB state across container restarts (see `doc/DOCKER.md`).

doc/DEVELOPING.md

@@ -15,6 +15,14 @@ Current implementation status:
 - Node.js 20+
 - pnpm 9+
 
+## Dependency Lockfile Policy
+
+GitHub Actions owns `pnpm-lock.yaml`.
+
+- Do not commit `pnpm-lock.yaml` in pull requests.
+- Pull request CI validates dependency resolution when manifests change.
+- Pushes to `master` regenerate `pnpm-lock.yaml` with `pnpm install --lockfile-only --no-frozen-lockfile`, commit it back if needed, and then run verification with `--frozen-lockfile`.
+
 ## Start Dev
 
 From repo root:
@@ -29,6 +37,10 @@ This starts:
 - API server: `http://localhost:3100`
 - UI: served by the API server in dev middleware mode (same origin as API)
 
+`pnpm dev` runs the server in watch mode and restarts on changes from workspace packages (including adapter packages). Use `pnpm dev:once` to run without file watching.
+
+`pnpm dev:once` now tracks backend-relevant file changes and pending migrations. When the current boot is stale, the board UI shows a `Restart required` banner. You can also enable guarded auto-restart in `Instance Settings > Experimental`, which waits for queued/running local agent runs to finish before restarting the dev server.
+
 Tailscale/private-auth dev mode:
 
 ```sh
@@ -79,6 +91,10 @@ docker compose -f docker-compose.quickstart.yml up --build
 
 See `doc/DOCKER.md` for API key wiring (`OPENAI_API_KEY` / `ANTHROPIC_API_KEY`) and persistence details.
 
+## Docker For Untrusted PR Review
+
+For a separate review-oriented container that keeps `codex`/`claude` login state in Docker volumes and checks out PRs into an isolated scratch workspace, see `doc/UNTRUSTED-PR-REVIEW.md`.
+
 ## Database in Dev (Auto-Handled)
 
 For local development, leave `DATABASE_URL` unset.
@@ -114,6 +130,123 @@ When a local agent run has no resolved project/session workspace, Paperclip fall
 
 This path honors `PAPERCLIP_HOME` and `PAPERCLIP_INSTANCE_ID` in non-default setups.
 
+For `codex_local`, Paperclip also manages a per-company Codex home under the instance root and seeds it from the shared Codex login/config home (`$CODEX_HOME` or `~/.codex`):
+
+- `~/.paperclip/instances/default/companies/<company-id>/codex-home`
+
+## Worktree-local Instances
+
+When developing from multiple git worktrees, do not point two Paperclip servers at the same embedded PostgreSQL data directory.
+
+Instead, create a repo-local Paperclip config plus an isolated instance for the worktree:
+
+```sh
+paperclipai worktree init
+# or create the git worktree and initialize it in one step:
+pnpm paperclipai worktree:make paperclip-pr-432
+```
+
+This command:
+
+- writes repo-local files at `.paperclip/config.json` and `.paperclip/.env`
+- creates an isolated instance under `~/.paperclip-worktrees/instances/<worktree-id>/`
+- when run inside a linked git worktree, mirrors the effective git hooks into that worktree's private git dir
+- picks a free app port and embedded PostgreSQL port
+- by default seeds the isolated DB in `minimal` mode from the current effective Paperclip instance/config (repo-local worktree config when present, otherwise the default instance) via a logical SQL snapshot
+
+Seed modes:
+
+- `minimal` keeps core app state like companies, projects, issues, comments, approvals, and auth state, preserves schema for all tables, but omits row data from heavy operational history such as heartbeat runs, wake requests, activity logs, runtime services, and agent session state
+- `full` makes a full logical clone of the source instance
+- `--no-seed` creates an empty isolated instance
+
+After `worktree init`, both the server and the CLI auto-load the repo-local `.paperclip/.env` when run inside that worktree, so normal commands like `pnpm dev`, `paperclipai doctor`, and `paperclipai db:backup` stay scoped to the worktree instance.
+
+That repo-local env also sets:
+
+- `PAPERCLIP_IN_WORKTREE=true`
+- `PAPERCLIP_WORKTREE_NAME=<worktree-name>`
+- `PAPERCLIP_WORKTREE_COLOR=<hex-color>`
+
+The server/UI use those values for worktree-specific branding such as the top banner and dynamically colored favicon.
+
+Print shell exports explicitly when needed:
+
+```sh
+paperclipai worktree env
+# or:
+eval "$(paperclipai worktree env)"
+```
+
+### Worktree CLI Reference
+
+**`pnpm paperclipai worktree init [options]`** — Create repo-local config/env and an isolated instance for the current worktree.
+
+| Option | Description |
+|---|---|
+| `--name <name>` | Display name used to derive the instance id |
+| `--instance <id>` | Explicit isolated instance id |
+| `--home <path>` | Home root for worktree instances (default: `~/.paperclip-worktrees`) |
+| `--from-config <path>` | Source config.json to seed from |
+| `--from-data-dir <path>` | Source PAPERCLIP_HOME used when deriving the source config |
+| `--from-instance <id>` | Source instance id (default: `default`) |
+| `--server-port <port>` | Preferred server port |
+| `--db-port <port>` | Preferred embedded Postgres port |
+| `--seed-mode <mode>` | Seed profile: `minimal` or `full` (default: `minimal`) |
+| `--no-seed` | Skip database seeding from the source instance |
+| `--force` | Replace existing repo-local config and isolated instance data |
+
+Examples:
+
+```sh
+paperclipai worktree init --no-seed
+paperclipai worktree init --seed-mode full
+paperclipai worktree init --from-instance default
+paperclipai worktree init --from-data-dir ~/.paperclip
+paperclipai worktree init --force
+```
+
+**`pnpm paperclipai worktree:make <name> [options]`** — Create `~/NAME` as a git worktree, then initialize an isolated Paperclip instance inside it. This combines `git worktree add` with `worktree init` in a single step.
+
+| Option | Description |
+|---|---|
+| `--start-point <ref>` | Remote ref to base the new branch on (e.g. `origin/main`) |
+| `--instance <id>` | Explicit isolated instance id |
+| `--home <path>` | Home root for worktree instances (default: `~/.paperclip-worktrees`) |
+| `--from-config <path>` | Source config.json to seed from |
+| `--from-data-dir <path>` | Source PAPERCLIP_HOME used when deriving the source config |
+| `--from-instance <id>` | Source instance id (default: `default`) |
+| `--server-port <port>` | Preferred server port |
+| `--db-port <port>` | Preferred embedded Postgres port |
+| `--seed-mode <mode>` | Seed profile: `minimal` or `full` (default: `minimal`) |
+| `--no-seed` | Skip database seeding from the source instance |
+| `--force` | Replace existing repo-local config and isolated instance data |
+
+Examples:
+
+```sh
+pnpm paperclipai worktree:make paperclip-pr-432
+pnpm paperclipai worktree:make my-feature --start-point origin/main
+pnpm paperclipai worktree:make experiment --no-seed
+```
+
+**`pnpm paperclipai worktree env [options]`** — Print shell exports for the current worktree-local Paperclip instance.
+
+| Option | Description |
+|---|---|
+| `-c, --config <path>` | Path to config file |
+| `--json` | Print JSON instead of shell exports |
+
+Examples:
+
+```sh
+pnpm paperclipai worktree env
+pnpm paperclipai worktree env --json
+eval "$(pnpm paperclipai worktree env)"
+```
+
+For project execution worktrees, Paperclip can also run a project-defined provision command after it creates or reuses an isolated git worktree. Configure this on the project's execution workspace policy (`workspaceStrategy.provisionCommand`). The command runs inside the derived worktree and receives `PAPERCLIP_WORKSPACE_*`, `PAPERCLIP_PROJECT_ID`, `PAPERCLIP_AGENT_ID`, and `PAPERCLIP_ISSUE_*` environment variables so each repo can bootstrap itself however it wants.
+
 ## Quick Health Checks
 
 In another terminal:
@@ -141,6 +274,36 @@ pnpm dev
 
 If you set `DATABASE_URL`, the server will use that instead of embedded PostgreSQL.
 
+## Automatic DB Backups
+
+Paperclip can run automatic DB backups on a timer. Defaults:
+
+- enabled
+- every 60 minutes
+- retain 30 days
+- backup dir: `~/.paperclip/instances/default/data/backups`
+
+Configure these in:
+
+```sh
+pnpm paperclipai configure --section database
+```
+
+Run a one-off backup manually:
+
+```sh
+pnpm paperclipai db:backup
+# or:
+pnpm db:backup
+```
+
+Environment overrides:
+
+- `PAPERCLIP_DB_BACKUP_ENABLED=true|false`
+- `PAPERCLIP_DB_BACKUP_INTERVAL_MINUTES=<minutes>`
+- `PAPERCLIP_DB_BACKUP_RETENTION_DAYS=<days>`
+- `PAPERCLIP_DB_BACKUP_DIR=/absolute/or/~/path`
+
 ## Secrets in Dev
 
 Agent env vars now support secret references. By default, secret values are stored with local encryption and only secret refs are persisted in agent config.
@@ -216,5 +379,61 @@ Agent-oriented invite onboarding now exposes machine-readable API docs:
 
 - `GET /api/invites/:token` returns invite summary plus onboarding and skills index links.
 - `GET /api/invites/:token/onboarding` returns onboarding manifest details (registration endpoint, claim endpoint template, skill install hints).
+- `GET /api/invites/:token/onboarding.txt` returns a plain-text onboarding doc intended for both human operators and agents (llm.txt-style handoff), including optional inviter message and suggested network host candidates.
 - `GET /api/skills/index` lists available skill documents.
 - `GET /api/skills/paperclip` returns the Paperclip heartbeat skill markdown.
+
+## OpenClaw Join Smoke Test
+
+Run the end-to-end OpenClaw join smoke harness:
+
+```sh
+pnpm smoke:openclaw-join
+```
+
+What it validates:
+
+- invite creation for agent-only join
+- agent join request using `adapterType=openclaw`
+- board approval + one-time API key claim semantics
+- callback delivery on wakeup to a dockerized OpenClaw-style webhook receiver
+
+Required permissions:
+
+- This script performs board-governed actions (create invite, approve join, wakeup another agent).
+- In authenticated mode, run with board auth via `PAPERCLIP_AUTH_HEADER` or `PAPERCLIP_COOKIE`.
+
+Optional auth flags (for authenticated mode):
+
+- `PAPERCLIP_AUTH_HEADER` (for example `Bearer ...`)
+- `PAPERCLIP_COOKIE` (session cookie header value)
+
+## OpenClaw Docker UI One-Command Script
+
+To boot OpenClaw in Docker and print a host-browser dashboard URL in one command:
+
+```sh
+pnpm smoke:openclaw-docker-ui
+```
+
+This script lives at `scripts/smoke/openclaw-docker-ui.sh` and automates clone/build/config/start for Compose-based local OpenClaw UI testing.
+
+Pairing behavior for this smoke script:
+
+- default `OPENCLAW_DISABLE_DEVICE_AUTH=1` (no Control UI pairing prompt for local smoke; no extra pairing env vars required)
+- set `OPENCLAW_DISABLE_DEVICE_AUTH=0` to require standard device pairing
+
+Model behavior for this smoke script:
+
+- defaults to OpenAI models (`openai/gpt-5.2` + OpenAI fallback) so it does not require Anthropic auth by default
+
+State behavior for this smoke script:
+
+- defaults to isolated config dir `~/.openclaw-paperclip-smoke`
+- resets smoke agent state each run by default (`OPENCLAW_RESET_STATE=1`) to avoid stale provider/auth drift
+
+Networking behavior for this smoke script:
+
+- auto-detects and prints a Paperclip host URL reachable from inside OpenClaw Docker
+- default container-side host alias is `host.docker.internal` (override with `PAPERCLIP_HOST_FROM_CONTAINER` / `PAPERCLIP_HOST_PORT`)
+- if Paperclip rejects container hostnames in authenticated/private mode, allow `host.docker.internal` via `pnpm paperclipai allowed-hostname host.docker.internal` and restart Paperclip

doc/DOCKER.md

@@ -42,6 +42,32 @@ Optional overrides:
 PAPERCLIP_PORT=3200 PAPERCLIP_DATA_DIR=./data/pc docker compose -f docker-compose.quickstart.yml up --build
 ```
 
+If you change host port or use a non-local domain, set `PAPERCLIP_PUBLIC_URL` to the external URL you will use in browser/auth flows.
+
+## Authenticated Compose (Single Public URL)
+
+For authenticated deployments, set one canonical public URL and let Paperclip derive auth/callback defaults:
+
+```yaml
+services:
+  paperclip:
+    environment:
+      PAPERCLIP_DEPLOYMENT_MODE: authenticated
+      PAPERCLIP_DEPLOYMENT_EXPOSURE: private
+      PAPERCLIP_PUBLIC_URL: https://desk.koker.net
+```
+
+`PAPERCLIP_PUBLIC_URL` is used as the primary source for:
+
+- auth public base URL
+- Better Auth base URL defaults
+- bootstrap invite URL defaults
+- hostname allowlist defaults (hostname extracted from URL)
+
+Granular overrides remain available if needed (`PAPERCLIP_AUTH_PUBLIC_BASE_URL`, `BETTER_AUTH_URL`, `BETTER_AUTH_TRUSTED_ORIGINS`, `PAPERCLIP_ALLOWED_HOSTNAMES`).
+
+Set `PAPERCLIP_ALLOWED_HOSTNAMES` explicitly only when you need additional hostnames beyond the public URL host (for example Tailscale/LAN aliases or multiple private hostnames).
+
 ## Claude + Codex Local Adapters in Docker
 
 The image pre-installs:
@@ -67,12 +93,19 @@ Notes:
 - Without API keys, the app still runs normally.
 - Adapter environment checks in Paperclip will surface missing auth/CLI prerequisites.
 
+## Untrusted PR Review Container
+
+If you want a separate Docker environment for reviewing untrusted pull requests with `codex` or `claude`, use the dedicated review workflow in `doc/UNTRUSTED-PR-REVIEW.md`.
+
+That setup keeps CLI auth state in Docker volumes instead of your host home directory and uses a separate scratch workspace for PR checkouts and preview runs.
+
 ## Onboard Smoke Test (Ubuntu + npm only)
 
 Use this when you want to mimic a fresh machine that only has Ubuntu + npm and verify:
 
 - `npx paperclipai onboard --yes` completes
 - the server binds to `0.0.0.0:3100` so host access works
+- onboard/run banners and startup logs are visible in your terminal
 
 Build + run:
 
@@ -80,15 +113,24 @@ Build + run:
 ./scripts/docker-onboard-smoke.sh
 ```
 
-Open: `http://localhost:3100`
+Open: `http://localhost:3131` (default smoke host port)
 
 Useful overrides:
 
 ```sh
 HOST_PORT=3200 PAPERCLIPAI_VERSION=latest ./scripts/docker-onboard-smoke.sh
+PAPERCLIP_DEPLOYMENT_MODE=authenticated PAPERCLIP_DEPLOYMENT_EXPOSURE=private ./scripts/docker-onboard-smoke.sh
+SMOKE_DETACH=true SMOKE_METADATA_FILE=/tmp/paperclip-smoke.env PAPERCLIPAI_VERSION=latest ./scripts/docker-onboard-smoke.sh
 ```
 
 Notes:
 
 - Persistent data is mounted at `./data/docker-onboard-smoke` by default.
+- Container runtime user id defaults to your local `id -u` so the mounted data dir stays writable while avoiding root runtime.
+- Smoke script defaults to `authenticated/private` mode so `HOST=0.0.0.0` can be exposed to the host.
+- Smoke script defaults host port to `3131` to avoid conflicts with local Paperclip on `3100`.
+- Smoke script also defaults `PAPERCLIP_PUBLIC_URL` to `http://localhost:<HOST_PORT>` so bootstrap invite URLs and auth callbacks use the reachable host port instead of the container's internal `3100`.
+- In authenticated mode, the smoke script defaults `SMOKE_AUTO_BOOTSTRAP=true` and drives the real bootstrap path automatically: it signs up a real user, runs `paperclipai auth bootstrap-ceo` inside the container to mint a real bootstrap invite, accepts that invite over HTTP, and verifies board session access.
+- Run the script in the foreground to watch the onboarding flow; stop with `Ctrl+C` after validation.
+- Set `SMOKE_DETACH=true` to leave the container running for automation and optionally write shell-ready metadata to `SMOKE_METADATA_FILE`.
 - The image definition is in `Dockerfile.onboard-smoke`.

doc/OPENCLAW_ONBOARDING.md

@@ -0,0 +1,94 @@
+Use this exact checklist.
+
+1. Start Paperclip in auth mode.
+```bash
+cd <paperclip-repo-root>
+pnpm dev --tailscale-auth
+```
+Then verify:
+```bash
+curl -sS http://127.0.0.1:3100/api/health | jq
+```
+
+2. Start a clean/stock OpenClaw Docker.
+```bash
+OPENCLAW_RESET_STATE=1 OPENCLAW_BUILD=1 ./scripts/smoke/openclaw-docker-ui.sh
+```
+Open the printed `Dashboard URL` (includes `#token=...`) in your browser.
+
+3. In Paperclip UI, go to `http://127.0.0.1:3100/CLA/company/settings`.
+
+4. Use the OpenClaw invite prompt flow.
+- In the Invites section, click `Generate OpenClaw Invite Prompt`.
+- Copy the generated prompt from `OpenClaw Invite Prompt`.
+- Paste it into OpenClaw main chat as one message.
+- If it stalls, send one follow-up: `How is onboarding going? Continue setup now.`
+
+Security/control note:
+- The OpenClaw invite prompt is created from a controlled endpoint:
+  - `POST /api/companies/{companyId}/openclaw/invite-prompt`
+  - board users with invite permission can call it
+  - agent callers are limited to the company CEO agent
+
+5. Approve the join request in Paperclip UI, then confirm the OpenClaw agent appears in CLA agents.
+
+6. Gateway preflight (required before task tests).
+- Confirm the created agent uses `openclaw_gateway` (not `openclaw`).
+- Confirm gateway URL is `ws://...` or `wss://...`.
+- Confirm gateway token is non-trivial (not empty / not 1-char placeholder).
+- The OpenClaw Gateway adapter UI should not expose `disableDeviceAuth` for normal onboarding.
+- Confirm pairing mode is explicit:
+  - required default: device auth enabled (`adapterConfig.disableDeviceAuth` false/absent) with persisted `adapterConfig.devicePrivateKeyPem`
+  - do not rely on `disableDeviceAuth` for normal onboarding
+- If you can run API checks with board auth:
+```bash
+AGENT_ID="<newly-created-agent-id>"
+curl -sS -H "Cookie: $PAPERCLIP_COOKIE" "http://127.0.0.1:3100/api/agents/$AGENT_ID" | jq '{adapterType,adapterConfig:{url:.adapterConfig.url,tokenLen:(.adapterConfig.headers["x-openclaw-token"] // .adapterConfig.headers["x-openclaw-auth"] // "" | length),disableDeviceAuth:(.adapterConfig.disableDeviceAuth // false),hasDeviceKey:(.adapterConfig.devicePrivateKeyPem // "" | length > 0)}}'
+```
+- Expected: `adapterType=openclaw_gateway`, `tokenLen >= 16`, `hasDeviceKey=true`, and `disableDeviceAuth=false`.
+
+Pairing handshake note:
+- Clean run expectation: first task should succeed without manual pairing commands.
+- The adapter attempts one automatic pairing approval + retry on first `pairing required` (when shared gateway auth token/password is valid).
+- If auto-pair cannot complete (for example token mismatch or no pending request), the first gateway run may still return `pairing required`.
+- This is a separate approval from Paperclip invite approval. You must approve the pending device in OpenClaw itself.
+- Approve it in OpenClaw, then retry the task.
+- For local docker smoke, you can approve from host:
+```bash
+docker exec openclaw-docker-openclaw-gateway-1 sh -lc 'openclaw devices approve --latest --json --url "ws://127.0.0.1:18789" --token "$(node -p \"require(process.env.HOME+\\\"/.openclaw/openclaw.json\\\").gateway.auth.token\")"'
+```
+- You can inspect pending vs paired devices:
+```bash
+docker exec openclaw-docker-openclaw-gateway-1 sh -lc 'TOK="$(node -e \"const fs=require(\\\"fs\\\");const c=JSON.parse(fs.readFileSync(\\\"/home/node/.openclaw/openclaw.json\\\",\\\"utf8\\\"));process.stdout.write(c.gateway?.auth?.token||\\\"\\\");\")\"; openclaw devices list --json --url \"ws://127.0.0.1:18789\" --token \"$TOK\"'
+```
+
+7. Case A (manual issue test).
+- Create an issue assigned to the OpenClaw agent.
+- Put instructions: “post comment `OPENCLAW_CASE_A_OK_<timestamp>` and mark done.”
+- Verify in UI: issue status becomes `done` and comment exists.
+
+8. Case B (message tool test).
+- Create another issue assigned to OpenClaw.
+- Instructions: “send `OPENCLAW_CASE_B_OK_<timestamp>` to main webchat via message tool, then comment same marker on issue, then mark done.”
+- Verify both:
+  - marker comment on issue
+  - marker text appears in OpenClaw main chat
+
+9. Case C (new session memory/skills test).
+- In OpenClaw, start `/new` session.
+- Ask it to create a new CLA issue in Paperclip with unique title `OPENCLAW_CASE_C_CREATED_<timestamp>`.
+- Verify in Paperclip UI that new issue exists.
+
+10. Watch logs during test (optional but helpful):
+```bash
+docker compose -f /tmp/openclaw-docker/docker-compose.yml -f /tmp/openclaw-docker/.paperclip-openclaw.override.yml logs -f openclaw-gateway
+```
+
+11. Expected pass criteria.
+- Preflight: `openclaw_gateway` + non-placeholder token (`tokenLen >= 16`).
+- Pairing mode: stable `devicePrivateKeyPem` configured with device auth enabled (default path).
+- Case A: `done` + marker comment.
+- Case B: `done` + marker comment + main-chat message visible.
+- Case C: original task done and new issue created from `/new` session.
+
+If you want, I can also give you a single “observer mode” command that runs the stock smoke harness while you watch the same steps live in UI.

doc/PRODUCT.md

@@ -94,3 +94,53 @@ Canonical mode design and command expectations live in `doc/DEPLOYMENT-MODES.md`
 ## Further Detail
 
 See [SPEC.md](./SPEC.md) for the full technical specification and [TASKS.md](./TASKS.md) for the task management data model.
+
+---
+
+Paperclip’s core identity is a **control plane for autonomous AI companies**, centered on **companies, org charts, goals, issues/comments, heartbeats, budgets, approvals, and board governance**. The public docs are also explicit about the current boundaries: **tasks/comments are the built-in communication model**, Paperclip is **not a chatbot**, and it is **not a code review tool**. The roadmap already points toward **easier onboarding, cloud agents, easier agent configuration, plugins, better docs, and ClipMart/ClipHub-style reusable companies/templates**.
+
+## What Paperclip should do vs. not do
+
+**Do**
+
+- Stay **board-level and company-level**. Users should manage goals, orgs, budgets, approvals, and outputs.
+- Make the first five minutes feel magical: install, answer a few questions, see a CEO do something real.
+- Keep work anchored to **issues/comments/projects/goals**, even if the surface feels conversational.
+- Treat **agency / internal team / startup** as the same underlying abstraction with different templates and labels.
+- Make outputs first-class: files, docs, reports, previews, links, screenshots.
+- Provide **hooks into engineering workflows**: worktrees, preview servers, PR links, external review tools.
+- Use **plugins** for edge cases like rich chat, knowledge bases, doc editors, custom tracing.
+
+**Do not**
+
+- Do not make the core product a general chat app. The current product definition is explicitly task/comment-centric and “not a chatbot,” and that boundary is valuable.
+- Do not build a complete Jira/GitHub replacement. The repo/docs already position Paperclip as organization orchestration, not focused on pull-request review.
+- Do not build enterprise-grade RBAC first. The current V1 spec still treats multi-board governance and fine-grained human permissions as out of scope, so the first multi-user version should be coarse and company-scoped.
+- Do not lead with raw bash logs and transcripts. Default view should be human-readable intent/progress, with raw detail beneath.
+- Do not force users to understand provider/API-key plumbing unless absolutely necessary. There are active onboarding/auth issues already; friction here is clearly real.
+
+## Specific design goals
+
+1. **Time-to-first-success under 5 minutes**
+   A fresh user should go from install to “my CEO completed a first task” in one sitting.
+
+2. **Board-level abstraction always wins**
+   The default UI should answer: what is the company doing, who is doing it, why does it matter, what did it cost, and what needs my approval.
+
+3. **Conversation stays attached to work objects**
+   “Chat with CEO” should still resolve to strategy threads, decisions, tasks, or approvals.
+
+4. **Progressive disclosure**
+   Top layer: human-readable summary. Middle layer: checklist/steps/artifacts. Bottom layer: raw logs/tool calls/transcript.
+
+5. **Output-first**
+   Work is not done until the user can see the result: file, document, preview link, screenshot, plan, or PR.
+
+6. **Local-first, cloud-ready**
+   The mental model should not change between local solo use and shared/private or public/cloud deployment.
+
+7. **Safe autonomy**
+   Auto mode is allowed; hidden token burn is not.
+
+8. **Thin core, rich edges**
+   Put optional chat, knowledge, and special surfaces into plugins/extensions rather than bloating the control plane.

doc/PUBLISHING.md

@@ -1,196 +1,140 @@
 # Publishing to npm
 
-This document covers how to build and publish the `paperclipai` CLI package to npm.
+Low-level reference for how Paperclip packages are prepared and published to npm.
 
-## Prerequisites
+For the maintainer workflow, use [doc/RELEASING.md](RELEASING.md). This document focuses on packaging internals.
 
-- Node.js 20+
-- pnpm 9.15+
-- An npm account with publish access to the `paperclipai` package
-- Logged in to npm: `npm login`
+## Current Release Entry Points
 
-## One-Command Publish
+Use these scripts:
 
-The fastest way to publish — bumps version, builds, publishes, restores, commits, and tags in one shot:
+- [`scripts/release.sh`](../scripts/release.sh) for canary and stable publish flows
+- [`scripts/create-github-release.sh`](../scripts/create-github-release.sh) after pushing a stable tag
+- [`scripts/rollback-latest.sh`](../scripts/rollback-latest.sh) to repoint `latest`
+- [`scripts/build-npm.sh`](../scripts/build-npm.sh) for the CLI packaging build
+
+Paperclip no longer uses release branches or Changesets for publishing.
+
+## Why the CLI needs special packaging
+
+The CLI package, `paperclipai`, imports code from workspace packages such as:
+
+- `@paperclipai/server`
+- `@paperclipai/db`
+- `@paperclipai/shared`
+- adapter packages under `packages/adapters/`
+
+Those workspace references are valid in development but not in a publishable npm package. The release flow rewrites versions temporarily, then builds a publishable CLI bundle.
+
+## `build-npm.sh`
+
+Run:
 
 ```bash
-./scripts/bump-and-publish.sh patch          # 0.1.1 → 0.1.2
-./scripts/bump-and-publish.sh minor          # 0.1.1 → 0.2.0
-./scripts/bump-and-publish.sh major          # 0.1.1 → 1.0.0
-./scripts/bump-and-publish.sh 2.0.0          # set explicit version
-./scripts/bump-and-publish.sh patch --dry-run # everything except npm publish
-```
-
-The script runs all 6 steps below in order. It requires a clean working tree and an active `npm login` session (unless `--dry-run`). After it finishes, push:
-
-```bash
-git push && git push origin v<version>
-```
-
-## Manual Step-by-Step
-
-If you prefer to run each step individually:
-
-### Quick Reference
-
-```bash
-# Bump version
-./scripts/version-bump.sh patch      # 0.1.0 → 0.1.1
-
-# Build
 ./scripts/build-npm.sh
-
-# Preview what will be published
-cd cli && npm pack --dry-run
-
-# Publish
-cd cli && npm publish --access public
-
-# Restore dev package.json
-mv cli/package.dev.json cli/package.json
 ```
 
-## Step-by-Step
+This script:
 
-### 1. Bump the version
+1. runs the forbidden token check unless `--skip-checks` is supplied
+2. runs `pnpm -r typecheck`
+3. bundles the CLI entrypoint with esbuild into `cli/dist/index.js`
+4. verifies the bundled entrypoint with `node --check`
+5. rewrites `cli/package.json` into a publishable npm manifest and stores the dev copy as `cli/package.dev.json`
+6. copies the repo `README.md` into `cli/README.md` for npm metadata
 
-```bash
-./scripts/version-bump.sh <patch|minor|major|X.Y.Z>
-```
+After the release script exits, the dev manifest and temporary files are restored automatically.
 
-This updates the version in two places:
+## Package discovery and versioning
 
-- `cli/package.json` — the source of truth
-- `cli/src/index.ts` — the Commander `.version()` call
+Public packages are discovered from:
+
+- `packages/`
+- `server/`
+- `cli/`
+
+`ui/` is ignored because it is private.
+
+The version rewrite step now uses [`scripts/release-package-map.mjs`](../scripts/release-package-map.mjs), which:
+
+- finds all public packages
+- sorts them topologically by internal dependencies
+- rewrites each package version to the target release version
+- rewrites internal `workspace:*` dependency references to the exact target version
+- updates the CLI's displayed version string
+
+Those rewrites are temporary. The working tree is restored after publish or dry-run.
+
+## Version formats
+
+Paperclip uses calendar versions:
+
+- stable: `YYYY.MDD.P`
+- canary: `YYYY.MDD.P-canary.N`
 
 Examples:
 
-```bash
-./scripts/version-bump.sh patch    # 0.1.0 → 0.1.1
-./scripts/version-bump.sh minor    # 0.1.0 → 0.2.0
-./scripts/version-bump.sh major    # 0.1.0 → 1.0.0
-./scripts/version-bump.sh 1.2.3   # set explicit version
-```
+- stable: `2026.318.0`
+- canary: `2026.318.1-canary.2`
 
-### 2. Build
+## Publish model
+
+### Canary
+
+Canaries publish under the npm dist-tag `canary`.
+
+Example:
+
+- `paperclipai@2026.318.1-canary.2`
+
+This keeps the default install path unchanged while allowing explicit installs with:
 
 ```bash
-./scripts/build-npm.sh
+npx paperclipai@canary onboard
 ```
 
-The build script runs five steps:
+### Stable
 
-1. **Forbidden token check** — scans tracked files for tokens listed in `.git/hooks/forbidden-tokens.txt`. If the file is missing (e.g. on a contributor's machine), the check passes silently. The script never prints which tokens it's searching for.
-2. **TypeScript type-check** — runs `pnpm -r typecheck` across all workspace packages.
-3. **esbuild bundle** — bundles the CLI entry point (`cli/src/index.ts`) and all workspace package code (`@paperclipai/*`) into a single file at `cli/dist/index.js`. External npm dependencies (express, postgres, etc.) are kept as regular imports.
-4. **Generate publishable package.json** — replaces `cli/package.json` with a version that has real npm dependency ranges instead of `workspace:*` references (see [package.dev.json](#packagedevjson) below).
-5. **Summary** — prints the bundle size and next steps.
+Stable publishes use the npm dist-tag `latest`.
 
-To skip the forbidden token check (e.g. in CI without the token list):
+Example:
+
+- `paperclipai@2026.318.0`
+
+Stable publishes do not create a release commit. Instead:
+
+- package versions are rewritten temporarily
+- packages are published from the chosen source commit
+- git tag `vYYYY.MDD.P` points at that original commit
+
+## Trusted publishing
+
+The intended CI model is npm trusted publishing through GitHub OIDC.
+
+That means:
+
+- no long-lived `NPM_TOKEN` in repository secrets
+- GitHub Actions obtains short-lived publish credentials
+- trusted publisher rules are configured per workflow file
+
+See [doc/RELEASE-AUTOMATION-SETUP.md](RELEASE-AUTOMATION-SETUP.md) for the GitHub/npm setup steps.
+
+## Rollback model
+
+Rollback does not unpublish anything.
+
+It repoints the `latest` dist-tag to a prior stable version:
 
 ```bash
-./scripts/build-npm.sh --skip-checks
+./scripts/rollback-latest.sh 2026.318.0
 ```
 
-### 3. Preview (optional)
+This is the fastest way to restore the default install path if a stable release is bad.
 
-See what npm will publish:
+## Related Files
 
-```bash
-cd cli && npm pack --dry-run
-```
-
-### 4. Publish
-
-```bash
-cd cli && npm publish --access public
-```
-
-### 5. Restore dev package.json
-
-After publishing, restore the workspace-aware `package.json`:
-
-```bash
-mv cli/package.dev.json cli/package.json
-```
-
-### 6. Commit and tag
-
-```bash
-git add cli/package.json cli/src/index.ts
-git commit -m "chore: bump version to X.Y.Z"
-git tag vX.Y.Z
-```
-
-## package.dev.json
-
-During development, `cli/package.json` contains `workspace:*` references like:
-
-```json
-{
-  "dependencies": {
-    "@paperclipai/server": "workspace:*",
-    "@paperclipai/db": "workspace:*"
-  }
-}
-```
-
-These tell pnpm to resolve those packages from the local monorepo. This is great for development but **npm doesn't understand `workspace:*`** — publishing with these references would cause install failures for users.
-
-The build script solves this with a two-file swap:
-
-1. **Before building:** `cli/package.json` has `workspace:*` refs (the dev version).
-2. **During build (`build-npm.sh` step 4):**
-   - The dev `package.json` is copied to `package.dev.json` as a backup.
-   - `generate-npm-package-json.mjs` reads every workspace package's `package.json`, collects all their external npm dependencies, and writes a new `cli/package.json` with those real dependency ranges — no `workspace:*` refs.
-3. **After publishing:** you restore the dev version with `mv package.dev.json package.json`.
-
-The generated publishable `package.json` looks like:
-
-```json
-{
-  "name": "paperclipai",
-  "version": "0.1.0",
-  "bin": { "paperclipai": "./dist/index.js" },
-  "dependencies": {
-    "express": "^5.1.0",
-    "postgres": "^3.4.5",
-    "commander": "^13.1.0"
-  }
-}
-```
-
-`package.dev.json` is listed in `.gitignore` — it only exists temporarily on disk during the build/publish cycle.
-
-## How the bundle works
-
-The CLI is a monorepo package that imports code from `@paperclipai/server`, `@paperclipai/db`, `@paperclipai/shared`, and several adapter packages. These workspace packages don't exist on npm.
-
-**esbuild** bundles all workspace TypeScript code into a single `dist/index.js` file (~250kb). External npm packages (express, postgres, zod, etc.) are left as normal `import` statements — they get installed by npm when a user runs `npx paperclipai onboard`.
-
-The esbuild configuration lives at `cli/esbuild.config.mjs`. It automatically reads every workspace package's `package.json` to determine which dependencies are external (real npm packages) vs. internal (workspace code to bundle).
-
-## Forbidden token enforcement
-
-The build process includes the same forbidden-token check used by the git pre-commit hook. This catches any accidentally committed tokens before they reach npm.
-
-- Token list: `.git/hooks/forbidden-tokens.txt` (one token per line, `#` comments supported)
-- The file lives inside `.git/` and is never committed
-- If the file is missing, the check passes — contributors without the list can still build
-- The script never prints which tokens are being searched for
-- Matches are printed so you know which files to fix, but not which token triggered it
-
-Run the check standalone:
-
-```bash
-pnpm check:tokens
-```
-
-## npm scripts reference
-
-| Script | Command | Description |
-|---|---|---|
-| `bump-and-publish` | `pnpm bump-and-publish <type>` | One-command bump + build + publish + commit + tag |
-| `build:npm` | `pnpm build:npm` | Full build (check + typecheck + bundle + package.json) |
-| `version:bump` | `pnpm version:bump <type>` | Bump CLI version |
-| `check:tokens` | `pnpm check:tokens` | Run forbidden token check only |
+- [`scripts/build-npm.sh`](../scripts/build-npm.sh)
+- [`scripts/generate-npm-package-json.mjs`](../scripts/generate-npm-package-json.mjs)
+- [`scripts/release-package-map.mjs`](../scripts/release-package-map.mjs)
+- [`cli/esbuild.config.mjs`](../cli/esbuild.config.mjs)
+- [`doc/RELEASING.md`](RELEASING.md)

doc/RELEASE-AUTOMATION-SETUP.md

@@ -0,0 +1,281 @@
+# Release Automation Setup
+
+This document covers the GitHub and npm setup required for the current Paperclip release model:
+
+- automatic canaries from `master`
+- manual stable promotion from a chosen source ref
+- npm trusted publishing via GitHub OIDC
+- protected release infrastructure in a public repository
+
+Repo-side files that depend on this setup:
+
+- `.github/workflows/release.yml`
+- `.github/CODEOWNERS`
+
+Note:
+
+- the release workflows intentionally use `pnpm install --no-frozen-lockfile`
+- this matches the repo's current policy where `pnpm-lock.yaml` is refreshed by GitHub automation after manifest changes land on `master`
+- the publish jobs then restore `pnpm-lock.yaml` before running `scripts/release.sh`, so the release script still sees a clean worktree
+
+## 1. Merge the Repo Changes First
+
+Before touching GitHub or npm settings, merge the release automation code so the referenced workflow filenames already exist on the default branch.
+
+Required files:
+
+- `.github/workflows/release.yml`
+- `.github/CODEOWNERS`
+
+## 2. Configure npm Trusted Publishing
+
+Do this for every public package that Paperclip publishes.
+
+At minimum that includes:
+
+- `paperclipai`
+- `@paperclipai/server`
+- public packages under `packages/`
+
+### 2.1. In npm, open each package settings page
+
+For each package:
+
+1. open npm as an owner of the package
+2. go to the package settings / publishing access area
+3. add a trusted publisher for the GitHub repository `paperclipai/paperclip`
+
+### 2.2. Add one trusted publisher entry per package
+
+npm currently allows one trusted publisher configuration per package.
+
+Configure:
+
+- workflow: `.github/workflows/release.yml`
+
+Repository:
+
+- `paperclipai/paperclip`
+
+Environment name:
+
+- leave the npm trusted-publisher environment field blank
+
+Why:
+
+- the single `release.yml` workflow handles both canary and stable publishing
+- GitHub environments `npm-canary` and `npm-stable` still enforce different approval rules on the GitHub side
+
+### 2.3. Verify trusted publishing before removing old auth
+
+After the workflows are live:
+
+1. run a canary publish
+2. confirm npm publish succeeds without any `NPM_TOKEN`
+3. run a stable dry-run
+4. run one real stable publish
+
+Only after that should you remove old token-based access.
+
+## 3. Remove Legacy npm Tokens
+
+After trusted publishing works:
+
+1. revoke any repository or organization `NPM_TOKEN` secrets used for publish
+2. revoke any personal automation token that used to publish Paperclip
+3. if npm offers a package-level setting to restrict publishing to trusted publishers, enable it
+
+Goal:
+
+- no long-lived npm publishing token should remain in GitHub Actions
+
+## 4. Create GitHub Environments
+
+Create two environments in the GitHub repository:
+
+- `npm-canary`
+- `npm-stable`
+
+Path:
+
+1. GitHub repository
+2. `Settings`
+3. `Environments`
+4. `New environment`
+
+## 5. Configure `npm-canary`
+
+Recommended settings for `npm-canary`:
+
+- environment name: `npm-canary`
+- required reviewers: none
+- wait timer: none
+- deployment branches and tags:
+  - selected branches only
+  - allow `master`
+
+Reasoning:
+
+- every push to `master` should be able to publish a canary automatically
+- no human approval should be required for canaries
+
+## 6. Configure `npm-stable`
+
+Recommended settings for `npm-stable`:
+
+- environment name: `npm-stable`
+- required reviewers: at least one maintainer other than the person triggering the workflow when possible
+- prevent self-review: enabled
+- admin bypass: disabled if your team can tolerate it
+- wait timer: optional
+- deployment branches and tags:
+  - selected branches only
+  - allow `master`
+
+Reasoning:
+
+- stable publishes should require an explicit human approval gate
+- the workflow is manual, but the environment should still be the real control point
+
+## 7. Protect `master`
+
+Open the branch protection settings for `master`.
+
+Recommended rules:
+
+1. require pull requests before merging
+2. require status checks to pass before merging
+3. require review from code owners
+4. dismiss stale approvals when new commits are pushed
+5. restrict who can push directly to `master`
+
+At minimum, make sure workflow and release script changes cannot land without review.
+
+## 8. Enforce CODEOWNERS Review
+
+This repo now includes `.github/CODEOWNERS`, but GitHub only enforces it if branch protection requires code owner reviews.
+
+In branch protection for `master`, enable:
+
+- `Require review from Code Owners`
+
+Then verify the owner entries are correct for your actual maintainer set.
+
+Current file:
+
+- `.github/CODEOWNERS`
+
+If `@cryppadotta` is not the right reviewer identity in the public repo, change it before enabling enforcement.
+
+## 9. Protect Release Infrastructure Specifically
+
+These files should always trigger code owner review:
+
+- `.github/workflows/release.yml`
+- `scripts/release.sh`
+- `scripts/release-lib.sh`
+- `scripts/release-package-map.mjs`
+- `scripts/create-github-release.sh`
+- `scripts/rollback-latest.sh`
+- `doc/RELEASING.md`
+- `doc/PUBLISHING.md`
+
+If you want stronger controls, add a repository ruleset that explicitly blocks direct pushes to:
+
+- `.github/workflows/**`
+- `scripts/release*`
+
+## 10. Do Not Store a Claude Token in GitHub Actions
+
+Do not add a personal Claude or Anthropic token for automatic changelog generation.
+
+Recommended policy:
+
+- stable changelog generation happens locally from a trusted maintainer machine
+- canaries never generate changelogs
+
+This keeps LLM spending intentional and avoids a high-value token sitting in Actions.
+
+## 11. Verify the Canary Workflow
+
+After setup:
+
+1. merge a harmless commit to `master`
+2. open the `Release` workflow run triggered by that push
+3. confirm it passes verification
+4. confirm publish succeeds under the `npm-canary` environment
+5. confirm npm now shows a new `canary` release
+6. confirm a git tag named `canary/vYYYY.MDD.P-canary.N` was pushed
+
+Install-path check:
+
+```bash
+npx paperclipai@canary onboard
+```
+
+## 12. Verify the Stable Workflow
+
+After at least one good canary exists:
+
+1. resolve the target stable version with `./scripts/release.sh stable --date YYYY-MM-DD --print-version`
+2. prepare `releases/vYYYY.MDD.P.md` on the source commit you want to promote
+3. open `Actions` -> `Release`
+4. run it with:
+   - `source_ref`: the tested commit SHA or canary tag source commit
+   - `stable_date`: leave blank or set the intended UTC date like `2026-03-18`
+     do not enter a version like `2026.318.0`; the workflow computes that from the date
+   - `dry_run`: `true`
+5. confirm the dry-run succeeds
+6. rerun with `dry_run: false`
+7. approve the `npm-stable` environment when prompted
+8. confirm npm `latest` points to the new stable version
+9. confirm git tag `vYYYY.MDD.P` exists
+10. confirm the GitHub Release was created
+
+Implementation note:
+
+- the GitHub Actions stable workflow calls `create-github-release.sh` with `PUBLISH_REMOTE=origin`
+- local maintainer usage can still pass `PUBLISH_REMOTE=public-gh` explicitly when needed
+
+## 13. Suggested Maintainer Policy
+
+Use this policy going forward:
+
+- canaries are automatic and cheap
+- stables are manual and approved
+- only stables get public notes and announcements
+- release notes are committed before stable publish
+- rollback uses `npm dist-tag`, not unpublish
+
+## 14. Troubleshooting
+
+### Trusted publishing fails with an auth error
+
+Check:
+
+1. the workflow filename on GitHub exactly matches the filename configured in npm
+2. the package has the trusted publisher entry for the correct repository
+3. the job has `id-token: write`
+4. the job is running from the expected repository, not a fork
+
+### Stable workflow runs but never asks for approval
+
+Check:
+
+1. the `publish` job uses environment `npm-stable`
+2. the environment actually has required reviewers configured
+3. the workflow is running in the canonical repository, not a fork
+
+### CODEOWNERS does not trigger
+
+Check:
+
+1. `.github/CODEOWNERS` is on the default branch
+2. branch protection on `master` requires code owner review
+3. the owner identities in the file are valid reviewers with repository access
+
+## Related Docs
+
+- [doc/RELEASING.md](RELEASING.md)
+- [doc/PUBLISHING.md](PUBLISHING.md)
+- [doc/plans/2026-03-17-release-automation-and-versioning.md](plans/2026-03-17-release-automation-and-versioning.md)

doc/RELEASING.md

@@ -0,0 +1,251 @@
+# Releasing Paperclip
+
+Maintainer runbook for shipping Paperclip across npm, GitHub, and the website-facing changelog surface.
+
+The release model is now commit-driven:
+
+1. Every push to `master` publishes a canary automatically.
+2. Stable releases are manually promoted from a chosen tested commit or canary tag.
+3. Stable release notes live in `releases/vYYYY.MDD.P.md`.
+4. Only stable releases get GitHub Releases.
+
+## Versioning Model
+
+Paperclip uses calendar versions that still fit semver syntax:
+
+- stable: `YYYY.MDD.P`
+- canary: `YYYY.MDD.P-canary.N`
+
+Examples:
+
+- first stable on March 18, 2026: `2026.318.0`
+- second stable on March 18, 2026: `2026.318.1`
+- fourth canary for the `2026.318.1` line: `2026.318.1-canary.3`
+
+Important constraints:
+
+- the middle numeric slot is `MDD`, where `M` is the UTC month and `DD` is the zero-padded UTC day
+- use `2026.303.0` for March 3, not `2026.33.0`
+- do not use leading zeroes such as `2026.0318.0`
+- do not use four numeric segments such as `2026.3.18.1`
+- the semver-safe canary form is `2026.318.0-canary.1`
+
+## Release Surfaces
+
+Every stable release has four separate surfaces:
+
+1. **Verification** — the exact git SHA passes typecheck, tests, and build
+2. **npm** — `paperclipai` and public workspace packages are published
+3. **GitHub** — the stable release gets a git tag and GitHub Release
+4. **Website / announcements** — the stable changelog is published externally and announced
+
+A stable release is done only when all four surfaces are handled.
+
+Canaries only cover the first two surfaces plus an internal traceability tag.
+
+## Core Invariants
+
+- canaries publish from `master`
+- stables publish from an explicitly chosen source ref
+- tags point at the original source commit, not a generated release commit
+- stable notes are always `releases/vYYYY.MDD.P.md`
+- canaries never create GitHub Releases
+- canaries never require changelog generation
+
+## TL;DR
+
+### Canary
+
+Every push to `master` runs the canary path inside [`.github/workflows/release.yml`](../.github/workflows/release.yml).
+
+It:
+
+- verifies the pushed commit
+- computes the canary version for the current UTC date
+- publishes under npm dist-tag `canary`
+- creates a git tag `canary/vYYYY.MDD.P-canary.N`
+
+Users install canaries with:
+
+```bash
+npx paperclipai@canary onboard
+# or
+npx paperclipai@canary onboard --data-dir "$(mktemp -d /tmp/paperclip-canary.XXXXXX)"
+```
+
+### Stable
+
+Use [`.github/workflows/release.yml`](../.github/workflows/release.yml) from the Actions tab with the manual `workflow_dispatch` inputs.
+
+[Run the action here](https://github.com/paperclipai/paperclip/actions/workflows/release.yml)
+
+Inputs:
+
+- `source_ref`
+  - commit SHA, branch, or tag
+- `stable_date`
+  - optional UTC date override in `YYYY-MM-DD`
+  - enter a date like `2026-03-18`, not a version like `2026.318.0`
+- `dry_run`
+  - preview only when true
+
+Before running stable:
+
+1. pick the canary commit or tag you trust
+2. resolve the target stable version with `./scripts/release.sh stable --date "$(date +%F)" --print-version`
+3. create or update `releases/vYYYY.MDD.P.md` on that source ref
+4. run the stable workflow from that source ref
+
+Example:
+
+- `source_ref`: `master`
+- `stable_date`: `2026-03-18`
+- resulting stable version: `2026.318.0`
+
+The workflow:
+
+- re-verifies the exact source ref
+- computes the next stable patch slot for the chosen UTC date
+- publishes `YYYY.MDD.P` under npm dist-tag `latest`
+- creates git tag `vYYYY.MDD.P`
+- creates or updates the GitHub Release from `releases/vYYYY.MDD.P.md`
+
+## Local Commands
+
+### Preview a canary locally
+
+```bash
+./scripts/release.sh canary --dry-run
+```
+
+### Preview a stable locally
+
+```bash
+./scripts/release.sh stable --dry-run
+```
+
+### Publish a stable locally
+
+This is mainly for emergency/manual use. The normal path is the GitHub workflow.
+
+```bash
+./scripts/release.sh stable
+git push public-gh refs/tags/vYYYY.MDD.P
+PUBLISH_REMOTE=public-gh ./scripts/create-github-release.sh YYYY.MDD.P
+```
+
+## Stable Changelog Workflow
+
+Stable changelog files live at:
+
+- `releases/vYYYY.MDD.P.md`
+
+Canaries do not get changelog files.
+
+Recommended local generation flow:
+
+```bash
+VERSION="$(./scripts/release.sh stable --date 2026-03-18 --print-version)"
+claude --print --output-format stream-json --verbose --dangerously-skip-permissions --model claude-opus-4-6 "Use the release-changelog skill to draft or update releases/v${VERSION}.md for Paperclip. Read doc/RELEASING.md and .agents/skills/release-changelog/SKILL.md, then generate the stable changelog for v${VERSION} from commits since the last stable tag. Do not create a canary changelog."
+```
+
+The repo intentionally does not run this through GitHub Actions because:
+
+- canaries are too frequent
+- stable notes are the only public narrative surface that needs LLM help
+- maintainer LLM tokens should not live in Actions
+
+## Smoke Testing
+
+For a canary:
+
+```bash
+PAPERCLIPAI_VERSION=canary ./scripts/docker-onboard-smoke.sh
+```
+
+For the current stable:
+
+```bash
+PAPERCLIPAI_VERSION=latest ./scripts/docker-onboard-smoke.sh
+```
+
+Useful isolated variants:
+
+```bash
+HOST_PORT=3232 DATA_DIR=./data/release-smoke-canary PAPERCLIPAI_VERSION=canary ./scripts/docker-onboard-smoke.sh
+HOST_PORT=3233 DATA_DIR=./data/release-smoke-stable PAPERCLIPAI_VERSION=latest ./scripts/docker-onboard-smoke.sh
+```
+
+Automated browser smoke is also available:
+
+```bash
+gh workflow run release-smoke.yml -f paperclip_version=canary
+gh workflow run release-smoke.yml -f paperclip_version=latest
+```
+
+Minimum checks:
+
+- `npx paperclipai@canary onboard` installs
+- onboarding completes without crashes
+- authenticated login works with the smoke credentials
+- the browser lands in onboarding on a fresh instance
+- company creation succeeds
+- the first CEO agent is created
+- the first CEO heartbeat run is triggered
+
+## Rollback
+
+Rollback does not unpublish versions.
+
+It only moves the `latest` dist-tag back to a previous stable:
+
+```bash
+./scripts/rollback-latest.sh 2026.318.0 --dry-run
+./scripts/rollback-latest.sh 2026.318.0
+```
+
+Then fix forward with a new stable patch slot or release date.
+
+## Failure Playbooks
+
+### If the canary publishes but smoke testing fails
+
+Do not run stable.
+
+Instead:
+
+1. fix the issue on `master`
+2. merge the fix
+3. wait for the next automatic canary
+4. rerun smoke testing
+
+### If stable npm publish succeeds but tag push or GitHub release creation fails
+
+This is a partial release. npm is already live.
+
+Do this immediately:
+
+1. push the missing tag
+2. rerun `PUBLISH_REMOTE=public-gh ./scripts/create-github-release.sh YYYY.MDD.P`
+3. verify the GitHub Release notes point at `releases/vYYYY.MDD.P.md`
+
+Do not republish the same version.
+
+### If `latest` is broken after stable publish
+
+Roll back the dist-tag:
+
+```bash
+./scripts/rollback-latest.sh YYYY.MDD.P
+```
+
+Then fix forward with a new stable release.
+
+## Related Files
+
+- [`scripts/release.sh`](../scripts/release.sh)
+- [`scripts/release-package-map.mjs`](../scripts/release-package-map.mjs)
+- [`scripts/create-github-release.sh`](../scripts/create-github-release.sh)
+- [`scripts/rollback-latest.sh`](../scripts/rollback-latest.sh)
+- [`doc/PUBLISHING.md`](PUBLISHING.md)
+- [`doc/RELEASE-AUTOMATION-SETUP.md`](RELEASE-AUTOMATION-SETUP.md)

doc/SPEC-implementation.md

@@ -37,7 +37,7 @@ These decisions close open questions from `SPEC.md` for V1.
 | Visibility | Full visibility to board and all agents in same company |
 | Communication | Tasks + comments only (no separate chat system) |
 | Task ownership | Single assignee; atomic checkout required for `in_progress` transition |
-| Recovery | No automatic reassignment; stale work is surfaced, not silently fixed |
+| Recovery | No automatic reassignment; work recovery stays manual/explicit |
 | Agent adapters | Built-in `process` and `http` adapters |
 | Auth | Mode-dependent human auth (`local_trusted` implicit board in current code; authenticated mode uses sessions), API keys for agents |
 | Budget period | Monthly UTC calendar window |
@@ -106,7 +106,6 @@ A lightweight scheduler/worker in the server process handles:
 - heartbeat trigger checks
 - stuck run detection
 - budget threshold checks
-- stale task reporting generation
 
 Separate queue infrastructure is not required for V1.
 
@@ -331,6 +330,34 @@ Operational policy:
   - `asset_id` uuid fk not null
   - `issue_comment_id` uuid fk null
 
+## 7.15 `documents` + `document_revisions` + `issue_documents`
+
+- `documents` stores editable text-first documents:
+  - `id` uuid pk
+  - `company_id` uuid fk not null
+  - `title` text null
+  - `format` text not null (`markdown`)
+  - `latest_body` text not null
+  - `latest_revision_id` uuid null
+  - `latest_revision_number` int not null
+  - `created_by_agent_id` uuid fk null
+  - `created_by_user_id` uuid/text fk null
+  - `updated_by_agent_id` uuid fk null
+  - `updated_by_user_id` uuid/text fk null
+- `document_revisions` stores append-only history:
+  - `id` uuid pk
+  - `company_id` uuid fk not null
+  - `document_id` uuid fk not null
+  - `revision_number` int not null
+  - `body` text not null
+  - `change_summary` text null
+- `issue_documents` links documents to issues with a stable workflow key:
+  - `id` uuid pk
+  - `company_id` uuid fk not null
+  - `issue_id` uuid fk not null
+  - `document_id` uuid fk not null
+  - `key` text not null (`plan`, `design`, `notes`, etc.)
+
 ## 8. State Machines
 
 ## 8.1 Agent Status
@@ -414,6 +441,7 @@ All endpoints are under `/api` and return JSON.
 - `POST /companies`
 - `GET /companies/:companyId`
 - `PATCH /companies/:companyId`
+- `PATCH /companies/:companyId/branding`
 - `POST /companies/:companyId/archive`
 
 ## 10.2 Goals
@@ -442,6 +470,11 @@ All endpoints are under `/api` and return JSON.
 - `POST /companies/:companyId/issues`
 - `GET /issues/:issueId`
 - `PATCH /issues/:issueId`
+- `GET /issues/:issueId/documents`
+- `GET /issues/:issueId/documents/:key`
+- `PUT /issues/:issueId/documents/:key`
+- `GET /issues/:issueId/documents/:key/revisions`
+- `DELETE /issues/:issueId/documents/:key`
 - `POST /issues/:issueId/checkout`
 - `POST /issues/:issueId/release`
 - `POST /issues/:issueId/comments`
@@ -502,7 +535,6 @@ Dashboard payload must include:
 - open/in-progress/blocked/done issue counts
 - month-to-date spend and budget utilization
 - pending approvals count
-- stale task count
 
 ## 10.9 Error Semantics
 
@@ -681,7 +713,6 @@ Required UX behaviors:
 - global company selector
 - quick actions: pause/resume agent, create task, approve/reject request
 - conflict toasts on atomic checkout failure
-- clear stale-task indicators
 - no silent background failures; every failed run visible in UI
 
 ## 15. Operational Requirements
@@ -780,7 +811,6 @@ A release candidate is blocked unless these pass:
 
 - add company selector and org chart view
 - add approvals and cost pages
-- add operational dashboard and stale-task surfacing
 
 ## Milestone 6: Hardening and Release
 
@@ -814,20 +844,27 @@ V1 is complete only when all criteria are true:
 
 V1 supports company import/export using a portable package contract:
 
-- exactly one JSON entrypoint: `paperclip.manifest.json`
-- all other package files are markdown with frontmatter
-- agent convention:
-  - `agents/<slug>/AGENTS.md` (required for V1 export/import)
-  - `agents/<slug>/HEARTBEAT.md` (optional, import accepted)
-  - `agents/<slug>/*.md` (optional, import accepted)
+- markdown-first package rooted at `COMPANY.md`
+- implicit folder discovery by convention
+- `.paperclip.yaml` sidecar for Paperclip-specific fidelity
+- canonical base package is vendor-neutral and aligned with `docs/companies/companies-spec.md`
+- common conventions:
+  - `agents/<slug>/AGENTS.md`
+  - `teams/<slug>/TEAM.md`
+  - `projects/<slug>/PROJECT.md`
+  - `projects/<slug>/tasks/<slug>/TASK.md`
+  - `tasks/<slug>/TASK.md`
+  - `skills/<slug>/SKILL.md`
 
 Export/import behavior in V1:
 
-- export includes company metadata and/or agents based on selection
-- export strips environment-specific paths (`cwd`, local instruction file paths)
-- export never includes secret values; secret requirements are reported
+- export emits a clean vendor-neutral markdown package plus `.paperclip.yaml`
+- projects and starter tasks are opt-in export content rather than default package content
+- export strips environment-specific paths (`cwd`, local instruction file paths, inline prompt duplication)
+- export never includes secret values; env inputs are reported as portable declarations instead
 - import supports target modes:
   - create a new company
   - import into an existing company
 - import supports collision strategies: `rename`, `skip`, `replace`
 - import supports preview (dry-run) before apply
+- GitHub imports warn on unpinned refs instead of blocking

doc/SPEC.md

@@ -188,12 +188,15 @@ The heartbeat is a protocol, not a runtime. Paperclip defines how to initiate an
 
 Agent configuration includes an **adapter** that defines how Paperclip invokes the agent. Initial adapters:
 
-| Adapter   | Mechanism               | Example                                       |
-| --------- | ----------------------- | --------------------------------------------- |
-| `process` | Execute a child process | `python run_agent.py --agent-id {id}`         |
-| `http`    | Send an HTTP request    | `POST https://openclaw.example.com/hook/{id}` |
+| Adapter              | Mechanism               | Example                                       |
+| -------------------- | ----------------------- | --------------------------------------------- |
+| `process`            | Execute a child process | `python run_agent.py --agent-id {id}`         |
+| `http`               | Send an HTTP request    | `POST https://openclaw.example.com/hook/{id}` |
+| `openclaw_gateway`   | OpenClaw gateway API    | Managed OpenClaw agent via gateway             |
+| `gemini_local`       | Gemini CLI process      | Local Gemini CLI with sandbox and approval     |
+| `hermes_local`       | Hermes agent process    | Local Hermes agent                             |
 
-The `process` and `http` adapters ship as defaults. Additional adapters can be added via the plugin system (see Plugin / Extension Architecture).
+The `process` and `http` adapters ship as defaults. Additional adapters have been added for specific agent runtimes (see list above), and new adapter types can be registered via the plugin system (see Plugin / Extension Architecture).
 
 ### Adapter Interface
 
@@ -429,7 +432,7 @@ The core Paperclip system must be extensible. Features like knowledge bases, ext
 - **Agent Adapter plugins** — new Adapter types can be registered via the plugin system
 - Plugin-registrable UI components (future)
 
-This isn't a V1 deliverable (we're not building a plugin framework upfront), but the architecture should not paint us into a corner. Keep boundaries clean so extensions are possible.
+The plugin framework has shipped. Plugins can register new adapter types, hook into lifecycle events, and contribute UI components (e.g. global toolbar buttons). A plugin SDK and CLI commands (`paperclipai plugin`) are available for authoring and installing plugins.
 
 ---
 

doc/UNTRUSTED-PR-REVIEW.md

@@ -0,0 +1,135 @@
+# Untrusted PR Review In Docker
+
+Use this workflow when you want Codex or Claude to inspect a pull request that you do not want touching your host machine directly.
+
+This is intentionally separate from the normal Paperclip dev image.
+
+## What this container isolates
+
+- `codex` auth/session state in a Docker volume, not your host `~/.codex`
+- `claude` auth/session state in a Docker volume, not your host `~/.claude`
+- `gh` auth state in the same container-local home volume
+- review clones, worktrees, dependency installs, and local databases in a writable scratch volume under `/work`
+
+By default this workflow does **not** mount your host repo checkout, your host home directory, or your SSH agent.
+
+## Files
+
+- `docker/untrusted-review/Dockerfile`
+- `docker-compose.untrusted-review.yml`
+- `review-checkout-pr` inside the container
+
+## Build and start a shell
+
+```sh
+docker compose -f docker-compose.untrusted-review.yml build
+docker compose -f docker-compose.untrusted-review.yml run --rm --service-ports review
+```
+
+That opens an interactive shell in the review container with:
+
+- Node + Corepack/pnpm
+- `codex`
+- `claude`
+- `gh`
+- `git`, `rg`, `fd`, `jq`
+
+## First-time login inside the container
+
+Run these once. The resulting login state persists in the `review-home` Docker volume.
+
+```sh
+gh auth login
+codex login
+claude login
+```
+
+If you prefer API-key auth instead of CLI login, pass keys through Compose env:
+
+```sh
+OPENAI_API_KEY=... ANTHROPIC_API_KEY=... docker compose -f docker-compose.untrusted-review.yml run --rm review
+```
+
+## Check out a PR safely
+
+Inside the container:
+
+```sh
+review-checkout-pr paperclipai/paperclip 432
+cd /work/checkouts/paperclipai-paperclip/pr-432
+```
+
+What this does:
+
+1. Creates or reuses a repo clone under `/work/repos/...`
+2. Fetches `pull/<pr>/head` from GitHub
+3. Creates a detached git worktree under `/work/checkouts/...`
+
+The checkout lives entirely inside the container volume.
+
+## Ask Codex or Claude to review it
+
+Inside the PR checkout:
+
+```sh
+codex
+```
+
+Then give it a prompt like:
+
+```text
+Review this PR as hostile input. Focus on security issues, data exfiltration paths, sandbox escapes, dangerous install/runtime scripts, auth changes, and subtle behavioral regressions. Do not modify files. Produce findings ordered by severity with file references.
+```
+
+Or with Claude:
+
+```sh
+claude
+```
+
+## Preview the Paperclip app from the PR
+
+Only do this when you intentionally want to execute the PR's code inside the container.
+
+Inside the PR checkout:
+
+```sh
+pnpm install
+HOST=0.0.0.0 pnpm dev
+```
+
+Open from the host:
+
+- `http://localhost:3100`
+
+The Compose file also exposes Vite's default port:
+
+- `http://localhost:5173`
+
+Notes:
+
+- `pnpm install` can run untrusted lifecycle scripts from the PR. That is why this happens inside the isolated container instead of on your host.
+- If you only want static inspection, do not run install/dev commands.
+- Paperclip's embedded PostgreSQL and local storage stay inside the container home volume via `PAPERCLIP_HOME=/home/reviewer/.paperclip-review`.
+
+## Reset state
+
+Remove the review container volumes when you want a clean environment:
+
+```sh
+docker compose -f docker-compose.untrusted-review.yml down -v
+```
+
+That deletes:
+
+- Codex/Claude/GitHub login state stored in `review-home`
+- cloned repos, worktrees, installs, and scratch data stored in `review-work`
+
+## Security limits
+
+This is a useful isolation boundary, but it is still Docker, not a full VM.
+
+- A reviewed PR can still access the container's network unless you disable it.
+- Any secrets you pass into the container are available to code you execute inside it.
+- Do not mount your host repo, host home, `.ssh`, or Docker socket unless you are intentionally weakening the boundary.
+- If you need a stronger boundary than this, use a disposable VM instead of Docker.

doc/experimental/issue-worktree-support.md

@@ -0,0 +1,62 @@
+# Issue worktree support
+
+Status: experimental, runtime-only, not shipping as a user-facing feature yet.
+
+This branch contains the runtime and seeding work needed for issue-scoped worktrees:
+
+- project execution workspace policy support
+- issue-level execution workspace settings
+- git worktree realization for isolated issue execution
+- optional command-based worktree provisioning
+- seeded worktree fixes for secrets key compatibility
+- seeded project workspace rebinding to the current git worktree
+
+We are intentionally not shipping the UI for this yet. The runtime code remains in place, but the main UI entrypoints are hard-gated off for now.
+
+## What works today
+
+- projects can carry execution workspace policy in the backend
+- issues can carry execution workspace settings in the backend
+- heartbeat execution can realize isolated git worktrees
+- runtime can run a project-defined provision command inside the derived worktree
+- seeded worktree instances can keep local-encrypted secrets working
+- seeded worktree instances can rebind same-repo project workspace paths onto the current git worktree
+
+## Hidden UI entrypoints
+
+These are the current user-facing UI surfaces for the feature, now intentionally disabled:
+
+- project settings:
+  - `ui/src/components/ProjectProperties.tsx`
+  - execution workspace policy controls
+  - git worktree base ref / branch template / parent dir
+  - provision / teardown command inputs
+
+- issue creation:
+  - `ui/src/components/NewIssueDialog.tsx`
+  - isolated issue checkout toggle
+  - defaulting issue execution workspace settings from project policy
+
+- issue editing:
+  - `ui/src/components/IssueProperties.tsx`
+  - issue-level workspace mode toggle
+  - defaulting issue execution workspace settings when project changes
+
+- agent/runtime settings:
+  - `ui/src/adapters/runtime-json-fields.tsx`
+  - runtime services JSON field, which is part of the broader workspace-runtime support surface
+
+## Why the UI is hidden
+
+- the runtime behavior is still being validated
+- the workflow and operator ergonomics are not final
+- we do not want to expose a partially-baked user-facing feature in issues, projects, or settings
+
+## Re-enable plan
+
+When this is ready to ship:
+
+- re-enable the gated UI sections in the files above
+- review wording and defaults for project and issue controls
+- decide which agent/runtime settings should remain advanced-only
+- add end-to-end product-level verification for the full UI workflow

doc/memory-landscape.md

@@ -0,0 +1,172 @@
+# Memory Landscape
+
+Date: 2026-03-17
+
+This document summarizes the memory systems referenced in task `PAP-530` and extracts the design patterns that matter for Paperclip.
+
+## What Paperclip Needs From This Survey
+
+Paperclip is not trying to become a single opinionated memory engine. The more useful target is a control-plane memory surface that:
+
+- stays company-scoped
+- lets each company choose a default memory provider
+- lets specific agents override that default
+- keeps provenance back to Paperclip runs, issues, comments, and documents
+- records memory-related cost and latency the same way the rest of the control plane records work
+- works with plugin-provided providers, not only built-ins
+
+The question is not "which memory project wins?" The question is "what is the smallest Paperclip contract that can sit above several very different memory systems without flattening away the useful differences?"
+
+## Quick Grouping
+
+### Hosted memory APIs
+
+- `mem0`
+- `supermemory`
+- `Memori`
+
+These optimize for a simple application integration story: send conversation/content plus an identity, then query for relevant memory or user context later.
+
+### Agent-centric memory frameworks / memory OSes
+
+- `MemOS`
+- `memU`
+- `EverMemOS`
+- `OpenViking`
+
+These treat memory as an agent runtime subsystem, not only as a search index. They usually add task memory, profiles, filesystem-style organization, async ingestion, or skill/resource management.
+
+### Local-first memory stores / indexes
+
+- `nuggets`
+- `memsearch`
+
+These emphasize local persistence, inspectability, and low operational overhead. They are useful because Paperclip is local-first today and needs at least one zero-config path.
+
+## Per-Project Notes
+
+| Project | Shape | Notable API / model | Strong fit for Paperclip | Main mismatch |
+|---|---|---|---|---|
+| [nuggets](https://github.com/NeoVertex1/nuggets) | local memory engine + messaging gateway | topic-scoped HRR memory with `remember`, `recall`, `forget`, fact promotion into `MEMORY.md` | good example of lightweight local memory and automatic promotion | very specific architecture; not a general multi-tenant service |
+| [mem0](https://github.com/mem0ai/mem0) | hosted + OSS SDK | `add`, `search`, `getAll`, `get`, `update`, `delete`, `deleteAll`; entity partitioning via `user_id`, `agent_id`, `run_id`, `app_id` | closest to a clean provider API with identities and metadata filters | provider owns extraction heavily; Paperclip should not assume every backend behaves like mem0 |
+| [MemOS](https://github.com/MemTensor/MemOS) | memory OS / framework | unified add-retrieve-edit-delete, memory cubes, multimodal memory, tool memory, async scheduler, feedback/correction | strong source for optional capabilities beyond plain search | much broader than the minimal contract Paperclip should standardize first |
+| [supermemory](https://github.com/supermemoryai/supermemory) | hosted memory + context API | `add`, `profile`, `search.memories`, `search.documents`, document upload, settings; automatic profile building and forgetting | strong example of "context bundle" rather than raw search results | heavily productized around its own ontology and hosted flow |
+| [memU](https://github.com/NevaMind-AI/memU) | proactive agent memory framework | file-system metaphor, proactive loop, intent prediction, always-on companion model | good source for when memory should trigger agent behavior, not just retrieval | proactive assistant framing is broader than Paperclip's task-centric control plane |
+| [Memori](https://github.com/MemoriLabs/Memori) | hosted memory fabric + SDK wrappers | registers against LLM SDKs, attribution via `entity_id` + `process_id`, sessions, cloud + BYODB | strong example of automatic capture around model clients | wrapper-centric design does not map 1:1 to Paperclip's run / issue / comment lifecycle |
+| [EverMemOS](https://github.com/EverMind-AI/EverMemOS) | conversational long-term memory system | MemCell extraction, structured narratives, user profiles, hybrid retrieval / reranking | useful model for provenance-rich structured memories and evolving profiles | focused on conversational memory rather than generalized control-plane events |
+| [memsearch](https://github.com/zilliztech/memsearch) | markdown-first local memory index | markdown as source of truth, `index`, `search`, `watch`, transcript parsing, plugin hooks | excellent baseline for a local built-in provider and inspectable provenance | intentionally simple; no hosted service semantics or rich correction workflow |
+| [OpenViking](https://github.com/volcengine/OpenViking) | context database | filesystem-style organization of memories/resources/skills, tiered loading, visualized retrieval trajectories | strong source for browse/inspect UX and context provenance | treats "context database" as a larger product surface than Paperclip should own |
+
+## Common Primitives Across The Landscape
+
+Even though the systems disagree on architecture, they converge on a few primitives:
+
+- `ingest`: add memory from text, messages, documents, or transcripts
+- `query`: search or retrieve memory given a task, question, or scope
+- `scope`: partition memory by user, agent, project, process, or session
+- `provenance`: carry enough metadata to explain where a memory came from
+- `maintenance`: update, forget, dedupe, compact, or correct memories over time
+- `context assembly`: turn raw memories into a prompt-ready bundle for the agent
+
+If Paperclip does not expose these, it will not adapt well to the systems above.
+
+## Where The Systems Differ
+
+These differences are exactly why Paperclip needs a layered contract instead of a single hard-coded engine.
+
+### 1. Who owns extraction?
+
+- `mem0`, `supermemory`, and `Memori` expect the provider to infer memories from conversations.
+- `memsearch` expects the host to decide what markdown to write, then indexes it.
+- `MemOS`, `memU`, `EverMemOS`, and `OpenViking` sit somewhere in between and often expose richer memory construction pipelines.
+
+Paperclip should support both:
+
+- provider-managed extraction
+- Paperclip-managed extraction with provider-managed storage / retrieval
+
+### 2. What is the source of truth?
+
+- `memsearch` and `nuggets` make the source inspectable on disk.
+- hosted APIs often make the provider store canonical.
+- filesystem-style systems like `OpenViking` and `memU` treat hierarchy itself as part of the memory model.
+
+Paperclip should not require a single storage shape. It should require normalized references back to Paperclip entities.
+
+### 3. Is memory just search, or also profile and planning state?
+
+- `mem0` and `memsearch` center search and CRUD.
+- `supermemory` adds user profiles as a first-class output.
+- `MemOS`, `memU`, `EverMemOS`, and `OpenViking` expand into tool traces, task memory, resources, and skills.
+
+Paperclip should make plain search the minimum contract and richer outputs optional capabilities.
+
+### 4. Is memory synchronous or asynchronous?
+
+- local tools often work synchronously in-process.
+- larger systems add schedulers, background indexing, compaction, or sync jobs.
+
+Paperclip needs both direct request/response operations and background maintenance hooks.
+
+## Paperclip-Specific Takeaways
+
+### Paperclip should own these concerns
+
+- binding a provider to a company and optionally overriding it per agent
+- mapping Paperclip entities into provider scopes
+- provenance back to issue comments, documents, runs, and activity
+- cost / token / latency reporting for memory work
+- browse and inspect surfaces in the Paperclip UI
+- governance on destructive operations
+
+### Providers should own these concerns
+
+- extraction heuristics
+- embedding / indexing strategy
+- ranking and reranking
+- profile synthesis
+- contradiction resolution and forgetting logic
+- storage engine details
+
+### The control-plane contract should stay small
+
+Paperclip does not need to standardize every feature from every provider. It needs:
+
+- a required portable core
+- optional capability flags for richer providers
+- a way to record provider-native ids and metadata without pretending all providers are equivalent internally
+
+## Recommended Direction
+
+Paperclip should adopt a two-layer memory model:
+
+1. `Memory binding + control plane layer`
+   Paperclip decides which provider key is in effect for a company, agent, or project, and it logs every memory operation with provenance and usage.
+
+2. `Provider adapter layer`
+   A built-in or plugin-supplied adapter turns Paperclip memory requests into provider-specific calls.
+
+The portable core should cover:
+
+- ingest / write
+- search / recall
+- browse / inspect
+- get by provider record handle
+- forget / correction
+- usage reporting
+
+Optional capabilities can cover:
+
+- profile synthesis
+- async ingestion
+- multimodal content
+- tool / resource / skill memory
+- provider-native graph browsing
+
+That is enough to support:
+
+- a local markdown-first baseline similar to `memsearch`
+- hosted services similar to `mem0`, `supermemory`, or `Memori`
+- richer agent-memory systems like `MemOS` or `OpenViking`
+
+without forcing Paperclip itself to become a monolithic memory engine.

doc/plans/2026-02-16-module-system.md

@@ -1,5 +1,7 @@
 # Paperclip Module System
 
+> Supersession note: the company-template/package-format direction in this document is no longer current. For the current markdown-first company import/export plan, see `doc/plans/2026-03-13-company-import-export-v2.md` and `docs/companies/companies-spec.md`.
+
 ## Overview
 
 Paperclip's module system lets you extend the control plane with new capabilities — revenue tracking, observability, notifications, dashboards — without forking core. Modules are self-contained packages that register routes, UI pages, database tables, and lifecycle hooks.

doc/plans/2026-03-11-agent-chat-ui-and-issue-backed-conversations.md

@@ -0,0 +1,329 @@
+# Agent Chat UI and Issue-Backed Conversations
+
+## Context
+
+`PAP-475` asks two related questions:
+
+1. What UI kit should Paperclip use if we add a chat surface with an agent?
+2. How should chat fit the product without breaking the current issue-centric model?
+
+This is not only a component-library decision. In Paperclip today:
+
+- V1 explicitly says communication is `tasks + comments only`, with no separate chat system.
+- Issues already carry assignment, audit trail, billing code, project linkage, goal linkage, and active run linkage.
+- Live run streaming already exists on issue detail pages.
+- Agent sessions already persist by `taskKey`, and today `taskKey` falls back to `issueId`.
+- The OpenClaw gateway adapter already supports an issue-scoped session key strategy.
+
+That means the cheapest useful path is not "add a second messaging product inside Paperclip." It is "add a better conversational UI on top of issue and run primitives we already have."
+
+## Current Constraints From the Codebase
+
+### Durable work object
+
+The durable object in Paperclip is the issue, not a chat thread.
+
+- `IssueDetail` already combines comments, linked runs, live runs, and activity into one timeline.
+- `CommentThread` already renders markdown comments and supports reply/reassignment flows.
+- `LiveRunWidget` already renders streaming assistant/tool/system output for active runs.
+
+### Session behavior
+
+Session continuity is already task-shaped.
+
+- `heartbeat.ts` derives `taskKey` from `taskKey`, then `taskId`, then `issueId`.
+- `agent_task_sessions` stores session state per company + agent + adapter + task key.
+- OpenClaw gateway supports `sessionKeyStrategy=issue|fixed|run`, and `issue` already matches the Paperclip mental model well.
+
+That means "chat with the CEO about this issue" naturally maps to one durable session per issue today without inventing a second session system.
+
+### Billing behavior
+
+Billing is already issue-aware.
+
+- `cost_events` can attach to `issueId`, `projectId`, `goalId`, and `billingCode`.
+- heartbeat context already propagates issue linkage into runs and cost rollups.
+
+If chat leaves the issue model, Paperclip would need a second billing story. That is avoidable.
+
+## UI Kit Recommendation
+
+## Recommendation: `assistant-ui`
+
+Use `assistant-ui` as the chat presentation layer.
+
+Why it fits Paperclip:
+
+- It is a real chat UI kit, not just a hook.
+- It is composable and aligned with shadcn-style primitives, which matches the current UI stack well.
+- It explicitly supports custom backends, which matters because Paperclip talks to agents through issue comments, heartbeats, and run streams rather than direct provider calls.
+- It gives us polished chat affordances quickly: message list, composer, streaming text, attachments, thread affordances, and markdown-oriented rendering.
+
+Why not make "the Vercel one" the primary choice:
+
+- Vercel AI SDK is stronger today than the older "just `useChat` over `/api/chat`" framing. Its transport layer is flexible and can support custom protocols.
+- But AI SDK is still better understood here as a transport/runtime protocol layer than as the best end-user chat surface for Paperclip.
+- Paperclip does not need Vercel to own message state, persistence, or the backend contract. Paperclip already has its own issue, run, and session model.
+
+So the clean split is:
+
+- `assistant-ui` for UI primitives
+- Paperclip-owned runtime/store for state, persistence, and transport
+- optional AI SDK usage later only if we want its stream protocol or client transport abstraction
+
+## Product Options
+
+### Option A: Separate chat object
+
+Create a new top-level chat/thread model unrelated to issues.
+
+Pros:
+
+- clean mental model if users want freeform conversation
+- easy to hide from issue boards
+
+Cons:
+
+- breaks the current V1 product decision that communication is issue-centric
+- needs new persistence, billing, session, permissions, activity, and wakeup rules
+- creates a second "why does this exist?" object beside issues
+- makes "pick up an old chat" a separate retrieval problem
+
+Verdict: not recommended for V1.
+
+### Option B: Every chat is an issue
+
+Treat chat as a UI mode over an issue. The issue remains the durable record.
+
+Pros:
+
+- matches current product spec
+- billing, runs, comments, approvals, and activity already work
+- sessions already resume on issue identity
+- works with all adapters, including OpenClaw, without new agent auth or a second API surface
+
+Cons:
+
+- some chats are not really "tasks" in a board sense
+- onboarding and review conversations may clutter normal issue lists
+
+Verdict: best V1 foundation.
+
+### Option C: Hybrid with hidden conversation issues
+
+Back every conversation with an issue, but allow a conversation-flavored issue mode that is hidden from default execution boards unless promoted.
+
+Pros:
+
+- preserves the issue-centric backend
+- gives onboarding/review chat a cleaner UX
+- preserves billing and session continuity
+
+Cons:
+
+- requires extra UI rules and possibly a small schema or filtering addition
+- can become a disguised second system if not kept narrow
+
+Verdict: likely the right product shape after a basic issue-backed MVP.
+
+## Recommended Product Model
+
+### Phase 1 product decision
+
+For the first implementation, chat should be issue-backed.
+
+More specifically:
+
+- the board opens a chat surface for an issue
+- sending a message is a comment mutation on that issue
+- the assigned agent is woken through the existing issue-comment flow
+- streaming output comes from the existing live run stream for that issue
+- durable assistant output remains comments and run history, not an extra transcript store
+
+This keeps Paperclip honest about what it is:
+
+- the control plane stays issue-centric
+- chat is a better way to interact with issue work, not a new collaboration product
+
+### Onboarding and CEO conversations
+
+For onboarding, weekly reviews, and "chat with the CEO", use a conversation issue rather than a global chat tab.
+
+Suggested shape:
+
+- create a board-initiated issue assigned to the CEO
+- mark it as conversation-flavored in UI treatment
+- optionally hide it from normal issue boards by default later
+- keep all cost/run/session linkage on that issue
+
+This solves several concerns at once:
+
+- no separate API key or direct provider wiring is needed
+- the same CEO adapter is used
+- old conversations are recovered through normal issue history
+- the CEO can still create or update real child issues from the conversation
+
+## Session Model
+
+### V1
+
+Use one durable conversation session per issue.
+
+That already matches current behavior:
+
+- adapter task sessions persist against `taskKey`
+- `taskKey` already falls back to `issueId`
+- OpenClaw already supports an issue-scoped session key
+
+This means "resume the CEO conversation later" works by reopening the same issue and waking the same agent on the same issue.
+
+### What not to add yet
+
+Do not add multi-thread-per-issue chat in the first pass.
+
+If Paperclip later needs several parallel threads on one issue, then add an explicit conversation identity and derive:
+
+- `taskKey = issue:<issueId>:conversation:<conversationId>`
+- OpenClaw `sessionKey = paperclip:conversation:<conversationId>`
+
+Until that requirement becomes real, one issue == one durable conversation is the simpler and better rule.
+
+## Billing Model
+
+Chat should not invent a separate billing pipeline.
+
+All chat cost should continue to roll up through the issue:
+
+- `cost_events.issueId`
+- project and goal rollups through existing relationships
+- issue `billingCode` when present
+
+If a conversation is important enough to exist, it is important enough to have a durable issue-backed audit and cost trail.
+
+This is another reason ephemeral freeform chat should not be the default.
+
+## UI Architecture
+
+### Recommended stack
+
+1. Keep Paperclip as the source of truth for message history and run state.
+2. Add `assistant-ui` as the rendering/composer layer.
+3. Build a Paperclip runtime adapter that maps:
+   - issue comments -> user/assistant messages
+   - live run deltas -> streaming assistant messages
+   - issue attachments -> chat attachments
+4. Keep current markdown rendering and code-block support where possible.
+
+### Interaction flow
+
+1. Board opens issue detail in "Chat" mode.
+2. Existing comment history is mapped into chat messages.
+3. When the board sends a message:
+   - `POST /api/issues/{id}/comments`
+   - optionally interrupt the active run if the UX wants "send and replace current response"
+4. Existing issue comment wakeup logic wakes the assignee.
+5. Existing `/issues/{id}/live-runs` and `/issues/{id}/active-run` data feeds drive streaming.
+6. When the run completes, durable state remains in comments/runs/activity as it does now.
+
+### Why this fits the current code
+
+Paperclip already has most of the backend pieces:
+
+- issue comments
+- run timeline
+- run log and event streaming
+- markdown rendering
+- attachment support
+- assignee wakeups on comments
+
+The missing piece is mostly the presentation and the mapping layer, not a new backend domain.
+
+## Agent Scope
+
+Do not launch this as "chat with every agent."
+
+Start narrower:
+
+- onboarding chat with CEO
+- workflow/review chat with CEO
+- maybe selected exec roles later
+
+Reasons:
+
+- it keeps the feature from becoming a second inbox/chat product
+- it limits permission and UX questions early
+- it matches the stated product demand
+
+If direct chat with other agents becomes useful later, the same issue-backed pattern can expand cleanly.
+
+## Recommended Delivery Phases
+
+### Phase 1: Chat UI on existing issues
+
+- add a chat presentation mode to issue detail
+- use `assistant-ui`
+- map comments + live runs into the chat surface
+- no schema change
+- no new API surface
+
+This is the highest-leverage step because it tests whether the UX is actually useful before product model expansion.
+
+### Phase 2: Conversation-flavored issues for CEO chat
+
+- add a lightweight conversation classification
+- support creation of CEO conversation issues from onboarding and workflow entry points
+- optionally hide these from normal backlog/board views by default
+
+The smallest implementation could be a label or issue metadata flag. If it becomes important enough, then promote it to a first-class issue subtype later.
+
+### Phase 3: Promotion and thread splitting only if needed
+
+Only if we later see a real need:
+
+- allow promoting a conversation to a formal task issue
+- allow several threads per issue with explicit conversation identity
+
+This should be demand-driven, not designed up front.
+
+## Clear Recommendation
+
+If the question is "what should we use?", the answer is:
+
+- use `assistant-ui` for the chat UI
+- do not treat raw Vercel AI SDK UI hooks as the main product answer
+- keep chat issue-backed in V1
+- use the current issue comment + run + session + billing model rather than inventing a parallel chat subsystem
+
+If the question is "how should we think about chat in Paperclip?", the answer is:
+
+- chat is a mode of interacting with issue-backed agent work
+- not a separate product silo
+- not an excuse to stop tracing work, cost, and session history back to the issue
+
+## Implementation Notes
+
+### Immediate implementation target
+
+The most defensible first build is:
+
+- add a chat tab or chat-focused layout on issue detail
+- back it with the currently assigned agent on that issue
+- use `assistant-ui` primitives over existing comments and live run events
+
+### Defer these until proven necessary
+
+- standalone global chat objects
+- multi-thread chat inside one issue
+- chat with every agent in the org
+- a second persistence layer for message history
+- separate cost tracking for chats
+
+## References
+
+- V1 communication model: `doc/SPEC-implementation.md`
+- Current issue/comment/run UI: `ui/src/pages/IssueDetail.tsx`, `ui/src/components/CommentThread.tsx`, `ui/src/components/LiveRunWidget.tsx`
+- Session persistence and task key derivation: `server/src/services/heartbeat.ts`, `packages/db/src/schema/agent_task_sessions.ts`
+- OpenClaw session routing: `packages/adapters/openclaw-gateway/README.md`
+- assistant-ui docs: <https://www.assistant-ui.com/docs>
+- assistant-ui repo: <https://github.com/assistant-ui/assistant-ui>
+- AI SDK transport docs: <https://ai-sdk.dev/docs/ai-sdk-ui/transport>

doc/plans/2026-03-13-TOKEN-OPTIMIZATION-PLAN.md

@@ -0,0 +1,397 @@
+# Token Optimization Plan
+
+Date: 2026-03-13  
+Related discussion: https://github.com/paperclipai/paperclip/discussions/449
+
+## Goal
+
+Reduce token consumption materially without reducing agent capability, control-plane visibility, or task completion quality.
+
+This plan is based on:
+
+- the current V1 control-plane design
+- the current adapter and heartbeat implementation
+- the linked user discussion
+- local runtime data from the default Paperclip instance on 2026-03-13
+
+## Executive Summary
+
+The discussion is directionally right about two things:
+
+1. We should preserve session and prompt-cache locality more aggressively.
+2. We should separate stable startup instructions from per-heartbeat dynamic context.
+
+But that is not enough on its own.
+
+After reviewing the code and local run data, the token problem appears to have four distinct causes:
+
+1. **Measurement inflation on sessioned adapters.** Some token counters, especially for `codex_local`, appear to be recorded as cumulative session totals instead of per-heartbeat deltas.
+2. **Avoidable session resets.** Task sessions are intentionally reset on timer wakes and manual wakes, which destroys cache locality for common heartbeat paths.
+3. **Repeated context reacquisition.** The `paperclip` skill tells agents to re-fetch assignments, issue details, ancestors, and full comment threads on every heartbeat. The API does not currently offer efficient delta-oriented alternatives.
+4. **Large static instruction surfaces.** Agent instruction files and globally injected skills are reintroduced at startup even when most of that content is unchanged and not needed for the current task.
+
+The correct approach is:
+
+1. fix telemetry so we can trust the numbers
+2. preserve reuse where it is safe
+3. make context retrieval incremental
+4. add session compaction/rotation so long-lived sessions do not become progressively more expensive
+
+## Validated Findings
+
+### 1. Token telemetry is at least partly overstated today
+
+Observed from the local default instance:
+
+- `heartbeat_runs`: 11,360 runs between 2026-02-18 and 2026-03-13
+- summed `usage_json.inputTokens`: `2,272,142,368,952`
+- summed `usage_json.cachedInputTokens`: `2,217,501,559,420`
+
+Those totals are not credible as true per-heartbeat usage for the observed prompt sizes.
+
+Supporting evidence:
+
+- `adapter.invoke.payload.prompt` averages were small:
+  - `codex_local`: ~193 chars average, 6,067 chars max
+  - `claude_local`: ~160 chars average, 1,160 chars max
+- despite that, many `codex_local` runs report millions of input tokens
+- one reused Codex session in local data spans 3,607 runs and recorded `inputTokens` growing up to `1,155,283,166`
+
+Interpretation:
+
+- for sessioned adapters, especially Codex, we are likely storing usage reported by the runtime as a **session total**, not a **per-run delta**
+- this makes trend reporting, optimization work, and customer trust worse
+
+This does **not** mean there is no real token problem. It means we need a trustworthy baseline before we can judge optimization impact.
+
+### 2. Timer wakes currently throw away reusable task sessions
+
+In `server/src/services/heartbeat.ts`, `shouldResetTaskSessionForWake(...)` returns `true` for:
+
+- `wakeReason === "issue_assigned"`
+- `wakeSource === "timer"`
+- manual on-demand wakes
+
+That means many normal heartbeats skip saved task-session resume even when the workspace is stable.
+
+Local data supports the impact:
+
+- `timer/system` runs: 6,587 total
+- only 976 had a previous session
+- only 963 ended with the same session
+
+So timer wakes are the largest heartbeat path and are mostly not resuming prior task state.
+
+### 3. We repeatedly ask agents to reload the same task context
+
+The `paperclip` skill currently tells agents to do this on essentially every heartbeat:
+
+- fetch assignments
+- fetch issue details
+- fetch ancestor chain
+- fetch full issue comments
+
+Current API shape reinforces that pattern:
+
+- `GET /api/issues/:id/comments` returns the full thread
+- there is no `since`, cursor, digest, or summary endpoint for heartbeat consumption
+- `GET /api/issues/:id` returns full enriched issue context, not a minimal delta payload
+
+This is safe but expensive. It forces the model to repeatedly consume unchanged information.
+
+### 4. Static instruction payloads are not separated cleanly from dynamic heartbeat prompts
+
+The user discussion suggested a bootstrap prompt. That is the right direction.
+
+Current state:
+
+- the UI exposes `bootstrapPromptTemplate`
+- adapter execution paths do not currently use it
+- several adapters prepend `instructionsFilePath` content directly into the per-run prompt or system prompt
+
+Result:
+
+- stable instructions are re-sent or re-applied in the same path as dynamic heartbeat content
+- we are not deliberately optimizing for provider prompt caching
+
+### 5. We inject more skill surface than most agents need
+
+Local adapters inject repo skills into runtime skill directories.
+
+Important `codex_local` nuance:
+
+- Codex does not read skills directly from the active worktree.
+- Paperclip discovers repo skills from the current checkout, then symlinks them into `$CODEX_HOME/skills` or `~/.codex/skills`.
+- If an existing Paperclip skill symlink already points at another live checkout, the current implementation skips it instead of repointing it.
+- This can leave Codex using stale skill content from a different worktree even after Paperclip-side skill changes land.
+- That is both a correctness risk and a token-analysis risk, because runtime behavior may not reflect the instructions in the checkout being tested.
+
+Current repo skill sizes:
+
+- `skills/paperclip/SKILL.md`: 17,441 bytes
+- `.agents/skills/create-agent-adapter/SKILL.md`: 31,832 bytes
+- `skills/paperclip-create-agent/SKILL.md`: 4,718 bytes
+- `skills/para-memory-files/SKILL.md`: 3,978 bytes
+
+That is nearly 58 KB of skill markdown before any company-specific instructions.
+
+Not all of that is necessarily loaded into model context every run, but it increases startup surface area and should be treated as a token budget concern.
+
+## Principles
+
+We should optimize tokens under these rules:
+
+1. **Do not lose functionality.** Agents must still be able to resume work safely, understand why tasks exist, and act within governance rules.
+2. **Prefer stable context over repeated context.** Unchanged instructions should not be resent through the most expensive path.
+3. **Prefer deltas over full reloads.** Heartbeats should consume only what changed since the last useful run.
+4. **Measure normalized deltas, not raw adapter claims.** Especially for sessioned CLIs.
+5. **Keep escape hatches.** Board/manual runs may still want a forced fresh session.
+
+## Plan
+
+## Phase 1: Make token telemetry trustworthy
+
+This should happen first.
+
+### Changes
+
+- Store both:
+  - raw adapter-reported usage
+  - Paperclip-normalized per-run usage
+- For sessioned adapters, compute normalized deltas against prior usage for the same persisted session.
+- Add explicit fields for:
+  - `sessionReused`
+  - `taskSessionReused`
+  - `promptChars`
+  - `instructionsChars`
+  - `hasInstructionsFile`
+  - `skillSetHash` or skill count
+  - `contextFetchMode` (`full`, `delta`, `summary`)
+- Add per-adapter parser tests that distinguish cumulative-session counters from per-run counters.
+
+### Why
+
+Without this, we cannot tell whether a reduction came from a real optimization or a reporting artifact.
+
+### Success criteria
+
+- per-run token totals stop exploding on long-lived sessions
+- a resumed session’s usage curve is believable and monotonic at the session level, but not double-counted at the run level
+- cost pages can show both raw and normalized numbers while we migrate
+
+## Phase 2: Preserve safe session reuse by default
+
+This is the highest-leverage behavior change.
+
+### Changes
+
+- Stop resetting task sessions on ordinary timer wakes.
+- Keep resetting on:
+  - explicit manual “fresh run” invocations
+  - assignment changes
+  - workspace mismatch
+  - model mismatch / invalid resume errors
+- Add an explicit wake flag like `forceFreshSession: true` when the board wants a reset.
+- Record why a session was reused or reset in run metadata.
+
+### Why
+
+Timer wakes are the dominant heartbeat path. Resetting them destroys both session continuity and prompt cache reuse.
+
+### Success criteria
+
+- timer wakes resume the prior task session in the large majority of stable-workspace cases
+- no increase in stale-session failures
+- lower normalized input tokens per timer heartbeat
+
+## Phase 3: Separate static bootstrap context from per-heartbeat context
+
+This is the right version of the discussion’s bootstrap idea.
+
+### Changes
+
+- Implement `bootstrapPromptTemplate` in adapter execution paths.
+- Use it only when starting a fresh session, not on resumed sessions.
+- Keep `promptTemplate` intentionally small and stable:
+  - who I am
+  - what triggered this wake
+  - which task/comment/approval to prioritize
+- Move long-lived setup text out of recurring per-run prompts where possible.
+- Add UI guidance and warnings when `promptTemplate` contains high-churn or large inline content.
+
+### Why
+
+Static instructions and dynamic wake context have different cache behavior and should be modeled separately.
+
+For `codex_local`, this also requires isolating the Codex skill home per worktree or teaching Paperclip to repoint its own skill symlinks when the source checkout changes. Otherwise prompt and skill improvements in the active worktree may not reach the running agent.
+
+### Success criteria
+
+- fresh-session prompts can remain richer without inflating every resumed heartbeat
+- resumed prompts become short and structurally stable
+- cache hit rates improve for session-preserving adapters
+
+## Phase 4: Make issue/task context incremental
+
+This is the biggest product change and likely the biggest real token saver after session reuse.
+
+### Changes
+
+Add heartbeat-oriented endpoints and skill behavior:
+
+- `GET /api/agents/me/inbox-lite`
+  - minimal assignment list
+  - issue id, identifier, status, priority, updatedAt, lastExternalCommentAt
+- `GET /api/issues/:id/heartbeat-context`
+  - compact issue state
+  - parent-chain summary
+  - latest execution summary
+  - change markers
+- `GET /api/issues/:id/comments?after=<cursor>` or `?since=<timestamp>`
+  - return only new comments
+- optional `GET /api/issues/:id/context-digest`
+  - server-generated compact summary for heartbeat use
+
+Update the `paperclip` skill so the default pattern becomes:
+
+1. fetch compact inbox
+2. fetch compact task context
+3. fetch only new comments unless this is the first read, a mention-triggered wake, or a cache miss
+4. fetch full thread only on demand
+
+### Why
+
+Today we are using full-fidelity board APIs as heartbeat APIs. That is convenient but token-inefficient.
+
+### Success criteria
+
+- after first task acquisition, most heartbeats consume only deltas
+- repeated blocked-task or long-thread work no longer replays the whole comment history
+- mention-triggered wakes still have enough context to respond correctly
+
+## Phase 5: Add session compaction and controlled rotation
+
+This protects against long-lived session bloat.
+
+### Changes
+
+- Add rotation thresholds per adapter/session:
+  - turns
+  - normalized input tokens
+  - age
+  - cache hit degradation
+- Before rotating, produce a structured carry-forward summary:
+  - current objective
+  - work completed
+  - open decisions
+  - blockers
+  - files/artifacts touched
+  - next recommended action
+- Persist that summary in task session state or runtime state.
+- Start the next session with:
+  - bootstrap prompt
+  - compact carry-forward summary
+  - current wake trigger
+
+### Why
+
+Even when reuse is desirable, some sessions become too expensive to keep alive indefinitely.
+
+### Success criteria
+
+- very long sessions stop growing without bound
+- rotating a session does not cause loss of task continuity
+- successful task completion rate stays flat or improves
+
+## Phase 6: Reduce unnecessary skill surface
+
+### Changes
+
+- Move from “inject all repo skills” to an allowlist per agent or per adapter.
+- Default local runtime skill set should likely be:
+  - `paperclip`
+- Add opt-in skills for specialized agents:
+  - `paperclip-create-agent`
+  - `para-memory-files`
+  - `create-agent-adapter`
+- Expose active skill set in agent config and run metadata.
+- For `codex_local`, either:
+  - run with a worktree-specific `CODEX_HOME`, or
+  - treat Paperclip-owned Codex skill symlinks as repairable when they point at a different checkout
+
+### Why
+
+Most agents do not need adapter-authoring or memory-system skills on every run.
+
+### Success criteria
+
+- smaller startup instruction surface
+- no loss of capability for specialist agents that explicitly need extra skills
+
+## Rollout Order
+
+Recommended order:
+
+1. telemetry normalization
+2. timer-wake session reuse
+3. bootstrap prompt implementation
+4. heartbeat delta APIs + `paperclip` skill rewrite
+5. session compaction/rotation
+6. skill allowlists
+
+## Acceptance Metrics
+
+We should treat this plan as successful only if we improve both efficiency and task outcomes.
+
+Primary metrics:
+
+- normalized input tokens per successful heartbeat
+- normalized input tokens per completed issue
+- cache-hit ratio for sessioned adapters
+- session reuse rate by invocation source
+- fraction of heartbeats that fetch full comment threads
+
+Guardrail metrics:
+
+- task completion rate
+- blocked-task rate
+- stale-session failure rate
+- manual intervention rate
+- issue reopen rate after agent completion
+
+Initial targets:
+
+- 30% to 50% reduction in normalized input tokens per successful resumed heartbeat
+- 80%+ session reuse on stable timer wakes
+- 80%+ reduction in full-thread comment reloads after first task read
+- no statistically meaningful regression in completion rate or failure rate
+
+## Concrete Engineering Tasks
+
+1. Add normalized usage fields and migration support for run analytics.
+2. Patch sessioned adapter accounting to compute deltas from prior session totals.
+3. Change `shouldResetTaskSessionForWake(...)` so timer wakes do not reset by default.
+4. Implement `bootstrapPromptTemplate` end-to-end in adapter execution.
+5. Add compact heartbeat context and incremental comment APIs.
+6. Rewrite `skills/paperclip/SKILL.md` around delta-fetch behavior.
+7. Add session rotation with carry-forward summaries.
+8. Replace global skill injection with explicit allowlists.
+9. Fix `codex_local` skill resolution so worktree-local skill changes reliably reach the runtime.
+
+## Recommendation
+
+Treat this as a two-track effort:
+
+- **Track A: correctness and no-regret wins**
+  - telemetry normalization
+  - timer-wake session reuse
+  - bootstrap prompt implementation
+- **Track B: structural token reduction**
+  - delta APIs
+  - skill rewrite
+  - session compaction
+  - skill allowlists
+
+If we only do Track A, we will improve things, but agents will still re-read too much unchanged task context.
+
+If we only do Track B without fixing telemetry first, we will not be able to prove the gains cleanly.

doc/plans/2026-03-13-agent-evals-framework.md

@@ -0,0 +1,775 @@
+# Agent Evals Framework Plan
+
+Date: 2026-03-13
+
+## Context
+
+We need evals for the thing Paperclip actually ships:
+
+- agent behavior produced by adapter config
+- prompt templates and bootstrap prompts
+- skill sets and skill instructions
+- model choice
+- runtime policy choices that affect outcomes and cost
+
+We do **not** primarily need a fine-tuning pipeline.
+We need a regression framework that can answer:
+
+- if we change prompts or skills, do agents still do the right thing?
+- if we switch models, what got better, worse, or more expensive?
+- if we optimize tokens, did we preserve task outcomes?
+- can we grow the suite over time from real Paperclip usage?
+
+This plan is based on:
+
+- `doc/GOAL.md`
+- `doc/PRODUCT.md`
+- `doc/SPEC-implementation.md`
+- `docs/agents-runtime.md`
+- `doc/plans/2026-03-13-TOKEN-OPTIMIZATION-PLAN.md`
+- Discussion #449: <https://github.com/paperclipai/paperclip/discussions/449>
+- OpenAI eval best practices: <https://developers.openai.com/api/docs/guides/evaluation-best-practices>
+- Promptfoo docs: <https://www.promptfoo.dev/docs/configuration/test-cases/> and <https://www.promptfoo.dev/docs/providers/custom-api/>
+- LangSmith complex agent eval docs: <https://docs.langchain.com/langsmith/evaluate-complex-agent>
+- Braintrust dataset/scorer docs: <https://www.braintrust.dev/docs/annotate/datasets> and <https://www.braintrust.dev/docs/evaluate/write-scorers>
+
+## Recommendation
+
+Paperclip should take a **two-stage approach**:
+
+1. **Start with Promptfoo now** for narrow, prompt-and-skill behavior evals across models.
+2. **Grow toward a first-party, repo-local eval harness in TypeScript** for full Paperclip scenario evals.
+
+So the recommendation is no longer “skip Promptfoo.” It is:
+
+- use Promptfoo as the fastest bootstrap layer
+- keep eval cases and fixtures in this repo
+- avoid making Promptfoo config the deepest long-term abstraction
+
+More specifically:
+
+1. The canonical eval definitions should live in this repo under a top-level `evals/` directory.
+2. `v0` should use Promptfoo to run focused test cases across models and providers.
+3. The longer-term harness should run **real Paperclip scenarios** against seeded companies/issues/agents, not just raw prompt completions.
+4. The scoring model should combine:
+   - deterministic checks
+   - structured rubric scoring
+   - pairwise candidate-vs-baseline judging
+   - efficiency metrics from normalized usage/cost telemetry
+5. The framework should compare **bundles**, not just models.
+
+A bundle is:
+
+- adapter type
+- model id
+- prompt template(s)
+- bootstrap prompt template
+- skill allowlist / skill content version
+- relevant runtime flags
+
+That is the right unit because that is what actually changes behavior in Paperclip.
+
+## Why This Is The Right Shape
+
+### 1. We need to evaluate system behavior, not only prompt output
+
+Prompt-only tools are useful, but Paperclip’s real failure modes are often:
+
+- wrong issue chosen
+- wrong API call sequence
+- bad delegation
+- failure to respect approval boundaries
+- stale session behavior
+- over-reading context
+- claiming completion without producing artifacts or comments
+
+Those are control-plane behaviors. They require scenario setup, execution, and trace inspection.
+
+### 2. The repo is already TypeScript-first
+
+The existing monorepo already uses:
+
+- `pnpm`
+- `tsx`
+- `vitest`
+- TypeScript across server, UI, shared contracts, and adapters
+
+A TypeScript-first harness will fit the repo and CI better than introducing a Python-first test subsystem as the default path.
+
+Python can stay optional later for specialty scorers or research experiments.
+
+### 3. We need provider/model comparison without vendor lock-in
+
+OpenAI’s guidance is directionally right:
+
+- eval early and often
+- use task-specific evals
+- log everything
+- prefer pairwise/comparison-style judging over open-ended scoring
+
+But OpenAI’s Evals API is not the right control plane for Paperclip as the primary system because our target is explicitly multi-model and multi-provider.
+
+### 4. Hosted eval products are useful, and Promptfoo is the right bootstrap tool
+
+The current tradeoff:
+
+- Promptfoo is very good for local, repo-based prompt/provider matrices and CI integration.
+- LangSmith is strong on trajectory-style agent evals.
+- Braintrust has a clean dataset + scorer + experiment model and strong TypeScript support.
+
+The community suggestion is directionally right:
+
+- Promptfoo lets us start small
+- it supports simple assertions like contains / not-contains / regex / custom JS
+- it can run the same cases across multiple models
+- it supports OpenRouter
+- it can move into CI later
+
+That makes it the best `v0` tool for “did this prompt/skill/model change obviously regress?”
+
+But Paperclip should still avoid making a hosted platform or a third-party config format the core abstraction before we have our own stable eval model.
+
+The right move is:
+
+- start with Promptfoo for quick wins
+- keep the data portable and repo-owned
+- build a thin first-party harness around Paperclip concepts as the system grows
+- optionally export to or integrate with other tools later if useful
+
+## What We Should Evaluate
+
+We should split evals into four layers.
+
+### Layer 1: Deterministic contract evals
+
+These should require no judge model.
+
+Examples:
+
+- agent comments on the assigned issue
+- no mutation outside the agent’s company
+- approval-required actions do not bypass approval flow
+- task transitions are legal
+- output contains required structured fields
+- artifact links exist when the task required an artifact
+- no full-thread refetch on delta-only cases once the API supports it
+
+These are cheap, reliable, and should be the first line of defense.
+
+### Layer 2: Single-step behavior evals
+
+These test narrow behaviors in isolation.
+
+Examples:
+
+- chooses the correct issue from inbox
+- writes a reasonable first status comment
+- decides to ask for approval instead of acting directly
+- delegates to the correct report
+- recognizes blocked state and reports it clearly
+
+These are the closest thing to prompt evals, but still framed in Paperclip terms.
+
+### Layer 3: End-to-end scenario evals
+
+These run a full heartbeat or short sequence of heartbeats against a seeded scenario.
+
+Examples:
+
+- new assignment pickup
+- long-thread continuation
+- mention-triggered clarification
+- approval-gated hire request
+- manager escalation
+- workspace coding task that must leave a meaningful issue update
+
+These should evaluate both final state and trace quality.
+
+### Layer 4: Efficiency and regression evals
+
+These are not “did the answer look good?” evals. They are “did we preserve quality while improving cost/latency?” evals.
+
+Examples:
+
+- normalized input tokens per successful heartbeat
+- normalized tokens per completed issue
+- session reuse rate
+- full-thread reload rate
+- wall-clock duration
+- cost per successful scenario
+
+This layer is especially important for token optimization work.
+
+## Core Design
+
+## 1. Canonical object: `EvalCase`
+
+Each eval case should define:
+
+- scenario setup
+- target bundle(s)
+- execution mode
+- expected invariants
+- scoring rubric
+- tags/metadata
+
+Suggested shape:
+
+```ts
+type EvalCase = {
+  id: string;
+  description: string;
+  tags: string[];
+  setup: {
+    fixture: string;
+    agentId: string;
+    trigger: "assignment" | "timer" | "on_demand" | "comment" | "approval";
+  };
+  inputs?: Record<string, unknown>;
+  checks: {
+    hard: HardCheck[];
+    rubric?: RubricCheck[];
+    pairwise?: PairwiseCheck[];
+  };
+  metrics: MetricSpec[];
+};
+```
+
+The important part is that the case is about a Paperclip scenario, not a standalone prompt string.
+
+## 2. Canonical object: `EvalBundle`
+
+Suggested shape:
+
+```ts
+type EvalBundle = {
+  id: string;
+  adapter: string;
+  model: string;
+  promptTemplate: string;
+  bootstrapPromptTemplate?: string;
+  skills: string[];
+  flags?: Record<string, string | number | boolean>;
+};
+```
+
+Every comparison run should say which bundle was tested.
+
+This avoids the common mistake of saying “model X is better” when the real change was model + prompt + skills + runtime behavior.
+
+## 3. Canonical output: `EvalTrace`
+
+We should capture a normalized trace for scoring:
+
+- run ids
+- prompts actually sent
+- session reuse metadata
+- issue mutations
+- comments created
+- approvals requested
+- artifacts created
+- token/cost telemetry
+- timing
+- raw outputs
+
+The scorer layer should never need to scrape ad hoc logs.
+
+## Scoring Framework
+
+## 1. Hard checks first
+
+Every eval should start with pass/fail checks that can invalidate the run immediately.
+
+Examples:
+
+- touched wrong company
+- skipped required approval
+- no issue update produced
+- returned malformed structured output
+- marked task done without required artifact
+
+If a hard check fails, the scenario fails regardless of style or judge score.
+
+## 2. Rubric scoring second
+
+Rubric scoring should use narrow criteria, not vague “how good was this?” prompts.
+
+Good rubric dimensions:
+
+- task understanding
+- governance compliance
+- useful progress communication
+- correct delegation
+- evidence of completion
+- concision / unnecessary verbosity
+
+Each rubric should be a small 0-1 or 0-2 decision, not a mushy 1-10 scale.
+
+## 3. Pairwise judging for candidate vs baseline
+
+OpenAI’s eval guidance is right that LLMs are better at discrimination than open-ended generation.
+
+So for non-deterministic quality checks, the default pattern should be:
+
+- run baseline bundle on the case
+- run candidate bundle on the same case
+- ask a judge model which is better on explicit criteria
+- allow `baseline`, `candidate`, or `tie`
+
+This is better than asking a judge for an absolute quality score with no anchor.
+
+## 4. Efficiency scoring is separate
+
+Do not bury efficiency inside a single blended quality score.
+
+Record it separately:
+
+- quality score
+- cost score
+- latency score
+
+Then compute a summary decision such as:
+
+- candidate is acceptable only if quality is non-inferior and efficiency is improved
+
+That is much easier to reason about than one magic number.
+
+## Suggested Decision Rule
+
+For PR gating:
+
+1. No hard-check regressions.
+2. No significant regression on required scenario pass rate.
+3. No significant regression on key rubric dimensions.
+4. If the change is token-optimization-oriented, require efficiency improvement on target scenarios.
+
+For deeper comparison reports, show:
+
+- pass rate
+- pairwise wins/losses/ties
+- median normalized tokens
+- median wall-clock time
+- cost deltas
+
+## Dataset Strategy
+
+We should explicitly build the dataset from three sources.
+
+### 1. Hand-authored seed cases
+
+Start here.
+
+These should cover core product invariants:
+
+- assignment pickup
+- status update
+- blocked reporting
+- delegation
+- approval request
+- cross-company access denial
+- issue comment follow-up
+
+These are small, clear, and stable.
+
+### 2. Production-derived cases
+
+Per OpenAI’s guidance, we should log everything and mine real usage for eval cases.
+
+Paperclip should grow eval coverage by promoting real runs into cases when we see:
+
+- regressions
+- interesting failures
+- edge cases
+- high-value success patterns worth preserving
+
+The initial version can be manual:
+
+- take a real run
+- redact/normalize it
+- convert it into an `EvalCase`
+
+Later we can automate trace-to-case generation.
+
+### 3. Adversarial and guardrail cases
+
+These should intentionally probe failure modes:
+
+- approval bypass attempts
+- wrong-company references
+- stale context traps
+- irrelevant long threads
+- misleading instructions in comments
+- verbosity traps
+
+This is where promptfoo-style red-team ideas can become useful later, but it is not the first slice.
+
+## Repo Layout
+
+Recommended initial layout:
+
+```text
+evals/
+  README.md
+  promptfoo/
+    promptfooconfig.yaml
+    prompts/
+    cases/
+  cases/
+    core/
+    approvals/
+    delegation/
+    efficiency/
+  fixtures/
+    companies/
+    issues/
+  bundles/
+    baseline/
+    experiments/
+  runners/
+    scenario-runner.ts
+    compare-runner.ts
+  scorers/
+    hard/
+    rubric/
+    pairwise/
+  judges/
+    rubric-judge.ts
+    pairwise-judge.ts
+  lib/
+    types.ts
+    traces.ts
+    metrics.ts
+  reports/
+    .gitignore
+```
+
+Why top-level `evals/`:
+
+- it makes evals feel first-class
+- it avoids hiding them inside `server/` even though they span adapters and runtime behavior
+- it leaves room for both TS and optional Python helpers later
+- it gives us a clean place for Promptfoo `v0` config plus the later first-party runner
+
+## Execution Model
+
+The harness should support three modes.
+
+### Mode A: Cheap local smoke
+
+Purpose:
+
+- run on PRs
+- keep cost low
+- catch obvious regressions
+
+Characteristics:
+
+- 5 to 20 cases
+- 1 or 2 bundles
+- mostly hard checks and narrow rubrics
+
+### Mode B: Candidate vs baseline compare
+
+Purpose:
+
+- evaluate a prompt/skill/model change before merge
+
+Characteristics:
+
+- paired runs
+- pairwise judging enabled
+- quality + efficiency diff report
+
+### Mode C: Nightly broader matrix
+
+Purpose:
+
+- compare multiple models and bundles
+- grow historical benchmark data
+
+Characteristics:
+
+- larger case set
+- multiple models
+- more expensive rubric/pairwise judging
+
+## CI and Developer Workflow
+
+Suggested commands:
+
+```sh
+pnpm evals:smoke
+pnpm evals:compare --baseline baseline/codex-default --candidate experiments/codex-lean-skillset
+pnpm evals:nightly
+```
+
+PR behavior:
+
+- run `evals:smoke` on prompt/skill/adapter/runtime changes
+- optionally trigger `evals:compare` for labeled PRs or manual runs
+
+Nightly behavior:
+
+- run larger matrix
+- save report artifact
+- surface trend lines on pass rate, pairwise wins, and efficiency
+
+## Framework Comparison
+
+## Promptfoo
+
+Best use for Paperclip:
+
+- prompt-level micro-evals
+- provider/model comparison
+- quick local CI integration
+- custom JS assertions and custom providers
+- bootstrap-layer evals for one skill or one agent workflow
+
+What changed in this recommendation:
+
+- Promptfoo is now the recommended **starting point**
+- especially for “one skill, a handful of cases, compare across models”
+
+Why it still should not be the only long-term system:
+
+- its primary abstraction is still prompt/provider/test-case oriented
+- Paperclip needs scenario setup, control-plane state inspection, and multi-step traces as first-class concepts
+
+Recommendation:
+
+- use Promptfoo first
+- store Promptfoo config and cases in-repo under `evals/promptfoo/`
+- use custom JS/TS assertions and, if needed later, a custom provider that calls Paperclip scenario runners
+- do not make Promptfoo YAML the only canonical Paperclip eval format once we outgrow prompt-level evals
+
+## LangSmith
+
+What it gets right:
+
+- final response evals
+- trajectory evals
+- single-step evals
+
+Why not the primary system today:
+
+- stronger fit for teams already centered on LangChain/LangGraph
+- introduces hosted/external workflow gravity before our own eval model is stable
+
+Recommendation:
+
+- copy the trajectory/final/single-step taxonomy
+- do not adopt the platform as the default requirement
+
+## Braintrust
+
+What it gets right:
+
+- TypeScript support
+- clean dataset/task/scorer model
+- production logging to datasets
+- experiment comparison over time
+
+Why not the primary system today:
+
+- still externalizes the canonical dataset and review workflow
+- we are not yet at the maturity where hosted experiment management should define the shape of the system
+
+Recommendation:
+
+- borrow its dataset/scorer/experiment mental model
+- revisit once we want hosted review and experiment history at scale
+
+## OpenAI Evals / Evals API
+
+What it gets right:
+
+- strong eval principles
+- emphasis on task-specific evals
+- continuous evaluation mindset
+
+Why not the primary system:
+
+- Paperclip must compare across models/providers
+- we do not want our primary eval runner coupled to one model vendor
+
+Recommendation:
+
+- use the guidance
+- do not use it as the core Paperclip eval runtime
+
+## First Implementation Slice
+
+The first version should be intentionally small.
+
+## Phase 0: Promptfoo bootstrap
+
+Build:
+
+- `evals/promptfoo/promptfooconfig.yaml`
+- 5 to 10 focused cases for one skill or one agent workflow
+- model matrix using the providers we care about most
+- mostly deterministic assertions:
+  - contains
+  - not-contains
+  - regex
+  - custom JS assertions
+
+Target scope:
+
+- one skill, or one narrow workflow such as assignment pickup / first status update
+- compare a small set of bundles across several models
+
+Success criteria:
+
+- we can run one command and compare outputs across models
+- prompt/skill regressions become visible quickly
+- the team gets signal before building heavier infrastructure
+
+## Phase 1: Skeleton and core cases
+
+Build:
+
+- `evals/` scaffold
+- `EvalCase`, `EvalBundle`, `EvalTrace` types
+- scenario runner for seeded local cases
+- 10 hand-authored core cases
+- hard checks only
+
+Target cases:
+
+- assigned issue pickup
+- write progress comment
+- ask for approval when required
+- respect company boundary
+- report blocked state
+- avoid marking done without artifact/comment evidence
+
+Success criteria:
+
+- a developer can run a local smoke suite
+- prompt/skill changes can fail the suite deterministically
+- Promptfoo `v0` cases either migrate into or coexist with this layer cleanly
+
+## Phase 2: Pairwise and rubric layer
+
+Build:
+
+- rubric scorer interface
+- pairwise judge runner
+- candidate vs baseline compare command
+- markdown/html report output
+
+Success criteria:
+
+- model/prompt bundle changes produce a readable diff report
+- we can tell “better”, “worse”, or “same” on curated scenarios
+
+## Phase 3: Efficiency integration
+
+Build:
+
+- normalized token/cost metrics into eval traces
+- cost and latency comparisons
+- efficiency gates for token optimization work
+
+Dependency:
+
+- this should align with the telemetry normalization work in `2026-03-13-TOKEN-OPTIMIZATION-PLAN.md`
+
+Success criteria:
+
+- quality and efficiency can be judged together
+- token-reduction work no longer relies on anecdotal improvements
+
+## Phase 4: Production-case ingestion
+
+Build:
+
+- tooling to promote real runs into new eval cases
+- metadata tagging
+- failure corpus growth process
+
+Success criteria:
+
+- the eval suite grows from real product behavior instead of staying synthetic
+
+## Initial Case Categories
+
+We should start with these categories:
+
+1. `core.assignment_pickup`
+2. `core.progress_update`
+3. `core.blocked_reporting`
+4. `governance.approval_required`
+5. `governance.company_boundary`
+6. `delegation.correct_report`
+7. `threads.long_context_followup`
+8. `efficiency.no_unnecessary_reloads`
+
+That is enough to start catching the classes of regressions we actually care about.
+
+## Important Guardrails
+
+### 1. Do not rely on judge models alone
+
+Every important scenario needs deterministic checks first.
+
+### 2. Do not gate PRs on a single noisy score
+
+Use pass/fail invariants plus a small number of stable rubric or pairwise checks.
+
+### 3. Do not confuse benchmark score with product quality
+
+The suite must keep growing from real runs, otherwise it will become a toy benchmark.
+
+### 4. Do not evaluate only final output
+
+Trajectory matters for agents:
+
+- did they call the right Paperclip APIs?
+- did they ask for approval?
+- did they communicate progress?
+- did they choose the right issue?
+
+### 5. Do not make the framework vendor-shaped
+
+Our eval model should survive changes in:
+
+- judge provider
+- candidate provider
+- adapter implementation
+- hosted tooling choices
+
+## Open Questions
+
+1. Should the first scenario runner invoke the real server over HTTP, or call services directly in-process?
+   My recommendation: start in-process for speed, then add HTTP-mode coverage once the model stabilizes.
+
+2. Should we support Python scorers in v1?
+   My recommendation: no. Keep v1 all-TypeScript.
+
+3. Should we commit baseline outputs?
+   My recommendation: commit case definitions and bundle definitions, but keep run artifacts out of git.
+
+4. Should we add hosted experiment tracking immediately?
+   My recommendation: no. Revisit after the local harness proves useful.
+
+## Final Recommendation
+
+Start with Promptfoo for immediate, narrow model-and-prompt comparisons, then grow into a first-party `evals/` framework in TypeScript that evaluates **Paperclip scenarios and bundles**, not just prompts.
+
+Use this structure:
+
+- Promptfoo for `v0` bootstrap
+- deterministic hard checks as the foundation
+- rubric and pairwise judging for non-deterministic quality
+- normalized efficiency metrics as a separate axis
+- repo-local datasets that grow from real runs
+
+Use external tools selectively:
+
+- Promptfoo as the initial path for narrow prompt/provider tests
+- Braintrust or LangSmith later if we want hosted experiment management
+
+But keep the canonical eval model inside the Paperclip repo and aligned to Paperclip’s actual control-plane behaviors.

doc/plans/2026-03-13-company-import-export-v2.md

@@ -0,0 +1,644 @@
+# 2026-03-13 Company Import / Export V2 Plan
+
+Status: Proposed implementation plan
+Date: 2026-03-13
+Audience: Product and engineering
+Supersedes for package-format direction:
+- `doc/plans/2026-02-16-module-system.md` sections that describe company templates as JSON-only
+- `docs/specs/cliphub-plan.md` assumptions about blueprint bundle shape where they conflict with the markdown-first package model
+
+## 1. Purpose
+
+This document defines the next-stage plan for Paperclip company import/export.
+
+The core shift is:
+
+- move from a Paperclip-specific JSON-first portability package toward a markdown-first package format
+- make GitHub repositories first-class package sources
+- treat the company package model as an extension of the existing Agent Skills ecosystem instead of inventing a separate skill format
+- support company, team, agent, and skill reuse without requiring a central registry
+
+The normative package format draft lives in:
+
+- `docs/companies/companies-spec.md`
+
+This plan is about implementation and rollout inside Paperclip.
+
+Adapter-wide skill rollout details live in:
+
+- `doc/plans/2026-03-14-adapter-skill-sync-rollout.md`
+
+## 2. Executive Summary
+
+Paperclip already has portability primitives in the repo:
+
+- server import/export/preview APIs
+- CLI import/export commands
+- shared portability types and validators
+
+Those primitives are being cut over to the new package model rather than extended for backward compatibility.
+
+The new direction is:
+
+1. markdown-first package authoring
+2. GitHub repo or local folder as the default source of truth
+3. a vendor-neutral base package spec for agent-company runtimes, not just Paperclip
+4. the company package model is explicitly an extension of Agent Skills
+5. no future dependency on `paperclip.manifest.json`
+6. implicit folder discovery by convention for the common case
+7. an always-emitted `.paperclip.yaml` sidecar for high-fidelity Paperclip-specific details
+8. package graph resolution at import time
+9. entity-level import UI with dependency-aware tree selection
+10. `skills.sh` compatibility is a V1 requirement for skill packages and skill installation flows
+11. adapter-aware skill sync surfaces so Paperclip can read, diff, enable, disable, and reconcile skills where the adapter supports it
+
+## 3. Product Goals
+
+### 3.1 Goals
+
+- A user can point Paperclip at a local folder or GitHub repo and import a company package without any registry.
+- A package is readable and writable by humans with normal git workflows.
+- A package can contain:
+  - company definition
+  - org subtree / team definition
+  - agent definitions
+  - optional starter projects and tasks
+  - reusable skills
+- V1 skill support is compatible with the existing `skills.sh` / Agent Skills ecosystem.
+- A user can import into:
+  - a new company
+  - an existing company
+- Import preview shows:
+  - what will be created
+  - what will be updated
+  - what is skipped
+  - what is referenced externally
+  - what needs secrets or approvals
+- Export preserves attribution, licensing, and pinned upstream references.
+- Export produces a clean vendor-neutral package plus a Paperclip sidecar.
+- `companies.sh` can later act as a discovery/index layer over repos implementing this format.
+
+### 3.2 Non-Goals
+
+- No central registry is required for package validity.
+- This is not full database backup/restore.
+- This does not attempt to export runtime state like:
+  - heartbeat runs
+  - API keys
+  - spend totals
+  - run sessions
+  - transient workspaces
+- This does not require a first-class runtime `teams` table before team portability ships.
+
+## 4. Current State In Repo
+
+Current implementation exists here:
+
+- shared types: `packages/shared/src/types/company-portability.ts`
+- shared validators: `packages/shared/src/validators/company-portability.ts`
+- server routes: `server/src/routes/companies.ts`
+- server service: `server/src/services/company-portability.ts`
+- CLI commands: `cli/src/commands/client/company.ts`
+
+Current product limitations:
+
+1. Import/export UX still needs deeper tree-selection and skill/package management polish.
+2. Adapter-specific skill sync remains uneven across adapters and must degrade cleanly when unsupported.
+3. Projects and starter tasks should stay opt-in on export rather than default package content.
+4. Import/export still needs stronger coverage around attribution, pin verification, and executable-package warnings.
+5. The current markdown frontmatter parser is intentionally lightweight and should stay constrained to the documented shape.
+
+## 5. Canonical Package Direction
+
+### 5.1 Canonical Authoring Format
+
+The canonical authoring format becomes a markdown-first package rooted in one of:
+
+- `COMPANY.md`
+- `TEAM.md`
+- `AGENTS.md`
+- `PROJECT.md`
+- `TASK.md`
+- `SKILL.md`
+
+The normative draft is:
+
+- `docs/companies/companies-spec.md`
+
+### 5.2 Relationship To Agent Skills
+
+Paperclip must not redefine `SKILL.md`.
+
+Rules:
+
+- `SKILL.md` stays Agent Skills compatible
+- the company package model is an extension of Agent Skills
+- the base package is vendor-neutral and intended for any agent-company runtime
+- Paperclip-specific fidelity lives in `.paperclip.yaml`
+- Paperclip may resolve and install `SKILL.md` packages, but it must not require a Paperclip-only skill format
+- `skills.sh` compatibility is a V1 requirement, not a future nice-to-have
+
+### 5.3 Agent-To-Skill Association
+
+`AGENTS.md` should associate skills by skill shortname or slug, not by verbose path in the common case.
+
+Preferred example:
+
+- `skills: [review, react-best-practices]`
+
+Resolution model:
+
+- `review` resolves to `skills/review/SKILL.md` by package convention
+- if the skill is external or referenced, the skill package owns that complexity
+- exporters should prefer shortname-based associations in `AGENTS.md`
+- importers should resolve the shortname against local package skills first, then referenced or installed company skills
+### 5.4 Base Package Vs Paperclip Extension
+
+The repo format should have two layers:
+
+- base package:
+  - minimal, readable, social, vendor-neutral
+  - implicit folder discovery by convention
+  - no Paperclip-only runtime fields by default
+- Paperclip extension:
+  - `.paperclip.yaml`
+  - adapter/runtime/permissions/budget/workspace fidelity
+  - emitted by Paperclip tools as a sidecar while the base package stays readable
+
+### 5.5 Relationship To Current V1 Manifest
+
+`paperclip.manifest.json` is not part of the future package direction.
+
+This should be treated as a hard cutover in product direction.
+
+- markdown-first repo layout is the target
+- no new work should deepen investment in the old manifest model
+- future portability APIs and UI should target the markdown-first model only
+
+## 6. Package Graph Model
+
+### 6.1 Entity Kinds
+
+Paperclip import/export should support these entity kinds:
+
+- company
+- team
+- agent
+- project
+- task
+- skill
+
+### 6.2 Team Semantics
+
+`team` is a package concept first, not a database-table requirement.
+
+In Paperclip V2 portability:
+
+- a team is an importable org subtree
+- it is rooted at a manager agent
+- it can be attached under a target manager in an existing company
+
+This avoids blocking portability on a future runtime `teams` model.
+
+Imported-team tracking should initially be package/provenance-based:
+
+- if a team package was imported, the imported agents should carry enough provenance to reconstruct that grouping
+- Paperclip can treat “this set of agents came from team package X” as the imported-team model
+- provenance grouping is the intended near- and medium-term team model for import/export
+- only add a first-class runtime `teams` table later if product needs move beyond what provenance grouping can express
+
+### 6.3 Dependency Graph
+
+Import should operate on an entity graph, not raw file selection.
+
+Examples:
+
+- selecting an agent auto-selects its required docs and skill refs
+- selecting a team auto-selects its subtree
+- selecting a company auto-selects all included entities by default
+- selecting a project auto-selects its starter tasks
+
+The preview output should reflect graph resolution explicitly.
+
+## 7. External References, Pinning, And Attribution
+
+### 7.1 Why This Matters
+
+Some packages will:
+
+- reference upstream files we do not want to republish
+- include third-party work where attribution must remain visible
+- need protection from branch hot-swapping
+
+### 7.2 Policy
+
+Paperclip should support source references in package metadata with:
+
+- repo
+- path
+- commit sha
+- optional blob sha
+- optional sha256
+- attribution
+- license
+- usage mode
+
+Usage modes:
+
+- `vendored`
+- `referenced`
+- `mirrored`
+
+Default exporter behavior for third-party content should be:
+
+- prefer `referenced`
+- preserve attribution
+- do not silently inline third-party content into exports
+
+### 7.3 Trust Model
+
+Imported package content should be classified by trust level:
+
+- markdown-only
+- markdown + assets
+- markdown + scripts/executables
+
+The UI and CLI should surface this clearly before apply.
+
+## 8. Import Behavior
+
+### 8.1 Supported Sources
+
+- local folder
+- local package root file
+- GitHub repo URL
+- GitHub subtree URL
+- direct URL to markdown/package root
+
+Registry-based discovery may be added later, but must remain optional.
+
+### 8.2 Import Targets
+
+- new company
+- existing company
+
+For existing company imports, the preview must support:
+
+- collision handling
+- attach-point selection for team imports
+- selective entity import
+
+### 8.3 Collision Strategy
+
+Current `rename | skip | replace` support remains, but matching should improve over time.
+
+Preferred matching order:
+
+1. prior install provenance
+2. stable package entity identity
+3. slug
+4. human name as weak fallback
+
+Slug-only matching is acceptable only as a transitional strategy.
+
+### 8.4 Required Preview Output
+
+Every import preview should surface:
+
+- target company action
+- entity-level create/update/skip plan
+- referenced external content
+- missing files
+- hash mismatch or pinning issues
+- env inputs, including required vs optional and default values when present
+- unsupported content types
+- trust/licensing warnings
+
+### 8.5 Adapter Skill Sync Surface
+
+People want skill management in the UI, but skills are adapter-dependent.
+
+That means portability and UI planning must include an adapter capability model for skills.
+
+Paperclip should define a new adapter surface area around skills:
+
+- list currently enabled skills for an agent
+- report how those skills are represented by the adapter
+- install or enable a skill
+- disable or remove a skill
+- report sync state between desired package config and actual adapter state
+
+Examples:
+
+- Claude Code / Codex style adapters may manage skills as local filesystem packages or adapter-owned skill directories
+- OpenClaw-style adapters may expose currently enabled skills through an API or a reflected config surface
+- some adapters may be read-only and only report what they have
+
+Planned adapter capability shape:
+
+- `supportsSkillRead`
+- `supportsSkillWrite`
+- `supportsSkillRemove`
+- `supportsSkillSync`
+- `skillStorageKind` such as `filesystem`, `remote_api`, `inline_config`, or `unknown`
+
+Baseline adapter interface:
+
+- `listSkills(agent)`
+- `applySkills(agent, desiredSkills)`
+- `removeSkill(agent, skillId)` optional
+- `getSkillSyncState(agent, desiredSkills)` optional
+
+Planned Paperclip behavior:
+
+- if an adapter supports read, Paperclip should show current skills in the UI
+- if an adapter supports write, Paperclip should let the user enable/disable imported skills
+- if an adapter supports sync, Paperclip should compute desired vs actual state and offer reconcile actions
+- if an adapter does not support these capabilities, the UI should still show the package-level desired skills but mark them unmanaged
+
+## 9. Export Behavior
+
+### 9.1 Default Export Target
+
+Default export target should become a markdown-first folder structure.
+
+Example:
+
+```text
+my-company/
+├── COMPANY.md
+├── agents/
+├── teams/
+└── skills/
+```
+
+### 9.2 Export Rules
+
+Exports should:
+
+- omit machine-local ids
+- omit timestamps and counters unless explicitly needed
+- omit secret values
+- omit local absolute paths
+- omit duplicated inline prompt content from `.paperclip.yaml` when `AGENTS.md` already carries the instructions
+- preserve references and attribution
+- emit `.paperclip.yaml` alongside the base package
+- express adapter env/secrets as portable env input declarations rather than exported secret binding ids
+- preserve compatible `SKILL.md` content as-is
+
+Projects and issues should not be exported by default.
+
+They should be opt-in through selectors such as:
+
+- `--projects project-shortname-1,project-shortname-2`
+- `--issues PAP-1,PAP-3`
+- `--project-issues project-shortname-1,project-shortname-2`
+
+This supports “clean public company package” workflows where a maintainer exports a follower-facing company package without bundling active work items every time.
+
+### 9.3 Export Units
+
+Initial export units:
+
+- company package
+- team package
+- single agent package
+
+Later optional units:
+
+- skill pack export
+- seed projects/tasks bundle
+
+## 10. Storage Model Inside Paperclip
+
+### 10.1 Short-Term
+
+In the first phase, imported entities can continue mapping onto current runtime tables:
+
+- company -> companies
+- agent -> agents
+- team -> imported agent subtree attachment plus package provenance grouping
+- skill -> company-scoped reusable package metadata plus agent-scoped desired-skill attachment state where supported
+
+### 10.2 Medium-Term
+
+Paperclip should add managed package/provenance records so imports are not anonymous one-off copies.
+
+Needed capabilities:
+
+- remember install origin
+- support re-import / upgrade
+- distinguish local edits from upstream package state
+- preserve external refs and package-level metadata
+- preserve imported team grouping without requiring a runtime `teams` table immediately
+- preserve desired-skill state separately from adapter runtime state
+- support both company-scoped reusable skills and agent-scoped skill attachments
+
+Suggested future tables:
+
+- package_installs
+- package_install_entities
+- package_sources
+- agent_skill_desires
+- adapter_skill_snapshots
+
+This is not required for phase 1 UI, but it is required for a robust long-term system.
+
+## 11. API Plan
+
+### 11.1 Keep Existing Endpoints Initially
+
+Retain:
+
+- `POST /api/companies/:companyId/export`
+- `POST /api/companies/import/preview`
+- `POST /api/companies/import`
+
+But evolve payloads toward the markdown-first graph model.
+
+### 11.2 New API Capabilities
+
+Add support for:
+
+- package root resolution from local/GitHub inputs
+- graph resolution preview
+- source pin and hash verification results
+- entity-level selection
+- team attach target selection
+- provenance-aware collision planning
+
+### 11.3 Parsing Changes
+
+Replace the current ad hoc markdown frontmatter parser with a real parser that can handle:
+
+- nested YAML
+- arrays/objects reliably
+- consistent round-tripping
+
+This is a prerequisite for the new package model.
+
+## 12. CLI Plan
+
+The CLI should continue to support direct import/export without a registry.
+
+Target commands:
+
+- `paperclipai company export <company-id> --out <path>`
+- `paperclipai company import --from <path-or-url> --dry-run`
+- `paperclipai company import --from <path-or-url> --target existing -C <company-id>`
+
+Planned additions:
+
+- `--package-kind company|team|agent`
+- `--attach-under <agent-id-or-slug>` for team imports
+- `--strict-pins`
+- `--allow-unpinned`
+- `--materialize-references`
+- `--sync-skills`
+
+## 13. UI Plan
+
+### 13.1 Company Settings Import / Export
+
+Add a real import/export section to Company Settings.
+
+Export UI:
+
+- export package kind selector
+- include options
+- local download/export destination guidance
+- attribution/reference summary
+
+Import UI:
+
+- source entry:
+  - upload/folder where supported
+  - GitHub URL
+  - generic URL
+- preview pane with:
+  - resolved package root
+  - dependency tree
+  - checkboxes by entity
+  - trust/licensing warnings
+  - secrets requirements
+  - collision plan
+
+### 13.2 Team Import UX
+
+If importing a team into an existing company:
+
+- show the subtree structure
+- require the user to choose where to attach it
+- preview manager/reporting updates before apply
+- preserve imported-team provenance so the UI can later say “these agents came from team package X”
+
+### 13.3 Skills UX
+
+See also:
+
+- `doc/plans/2026-03-14-skills-ui-product-plan.md`
+
+If importing skills:
+
+- show whether each skill is local, vendored, or referenced
+- show whether it contains scripts/assets
+- preserve Agent Skills compatibility in presentation and export
+- preserve `skills.sh` compatibility in both import and install flows
+- show agent skill attachments by shortname/slug rather than noisy file paths
+- treat agent skills as a dedicated agent tab, not just another subsection of configuration
+- show current adapter-reported skills when supported
+- show desired package skills separately from actual adapter state
+- offer reconcile actions when the adapter supports sync
+
+## 14. Rollout Phases
+
+### Phase 1: Stabilize Current V1 Portability
+
+- add tests for current portability flows
+- replace the frontmatter parser
+- add Company Settings UI for current import/export capabilities
+- start cutover work toward the markdown-first package reader
+
+### Phase 2: Markdown-First Package Reader
+
+- support `COMPANY.md` / `TEAM.md` / `AGENTS.md` root detection
+- build internal graph from markdown-first packages
+- support local folder and GitHub repo inputs natively
+- support agent skill references by shortname/slug
+- resolve local `skills/<slug>/SKILL.md` packages by convention
+- support `skills.sh`-compatible skill repos as V1 package sources
+
+### Phase 3: Graph-Based Import UX And Skill Surfaces
+
+- entity tree preview
+- checkbox selection
+- team subtree attach flow
+- licensing/trust/reference warnings
+- company skill library groundwork
+- dedicated agent `Skills` tab groundwork
+- adapter skill read/sync UI groundwork
+
+### Phase 4: New Export Model
+
+- export markdown-first folder structure by default
+
+### Phase 5: Provenance And Upgrades
+
+- persist install provenance
+- support package-aware re-import and upgrades
+- improve collision matching beyond slug-only
+- add imported-team provenance grouping
+- add desired-vs-actual skill sync state
+
+### Phase 6: Optional Seed Content
+
+- goals
+- projects
+- starter issues/tasks
+
+This phase is intentionally after the structural model is stable.
+
+## 15. Documentation Plan
+
+Primary docs:
+
+- `docs/companies/companies-spec.md` as the package-format draft
+- this implementation plan for rollout sequencing
+
+Docs to update later as implementation lands:
+
+- `doc/SPEC-implementation.md`
+- `docs/api/companies.md`
+- `docs/cli/control-plane-commands.md`
+- board operator docs for Company Settings import/export
+
+## 16. Open Questions
+
+1. Should imported skill packages be stored as managed package files in Paperclip storage, or only referenced at import time?
+   Decision: managed package files should support both company-scoped reuse and agent-scoped attachment.
+2. What is the minimum adapter skill interface needed to make the UI useful across Claude Code, Codex, OpenClaw, and future adapters?
+   Decision: use the baseline interface in section 8.5.
+3. Should Paperclip support direct local folder selection in the web UI, or keep that CLI-only initially?
+4. Do we want optional generated lock files in phase 2, or defer them until provenance work?
+5. How strict should pinning be by default for GitHub references:
+   - warn on unpinned
+   - or block in normal mode
+6. Is package-provenance grouping enough for imported teams, or do we expect product requirements soon that would justify a first-class runtime `teams` table?
+   Decision: provenance grouping is enough for the import/export product model for now.
+
+## 17. Recommendation
+
+Engineering should treat this as the current plan of record for company import/export beyond the existing V1 portability feature.
+
+Immediate next steps:
+
+1. accept `docs/companies/companies-spec.md` as the package-format draft
+2. implement phase 1 stabilization work
+3. build phase 2 markdown-first package reader before expanding ClipHub or `companies.sh`
+4. treat the old manifest-based format as deprecated and not part of the future surface
+
+This keeps Paperclip aligned with:
+
+- GitHub-native distribution
+- Agent Skills compatibility
+- a registry-optional ecosystem model

doc/plans/2026-03-13-features.md

@@ -0,0 +1,780 @@
+# Feature specs
+
+## 1) Guided onboarding + first-job magic
+
+The repo already has `onboard`, `doctor`, `run`, deployment modes, and even agent-oriented onboarding text/skills endpoints, but there are also current onboarding/auth validation issues and an open “onboard failed” report. That means this is not just polish; it is product-critical. ([GitHub][1])
+
+### Product decision
+
+Replace “configuration-first onboarding” with **interview-first onboarding**.
+
+### What we want
+
+- Ask 3–4 questions up front, not 20 settings.
+- Generate the right path automatically: local solo, shared private, or public cloud.
+- Detect what agent/runtime environment already exists.
+- Make it normal to have Claude/OpenClaw/Codex help complete setup.
+- End onboarding with a **real first task**, not a blank dashboard.
+
+### What we do not want
+
+- Provider jargon before value.
+- “Go find an API key” as the default first instruction.
+- A successful install that still leaves users unsure what to do next.
+
+### Proposed UX
+
+On first run, show an interview:
+
+```ts
+type OnboardingProfile = {
+  useCase: "startup" | "agency" | "internal_team";
+  companySource: "new" | "existing";
+  deployMode: "local_solo" | "shared_private" | "shared_public";
+  autonomyMode: "hands_on" | "hybrid" | "full_auto";
+  primaryRuntime: "claude_code" | "codex" | "openclaw" | "other";
+};
+```
+
+Questions:
+
+1. What are you building?
+2. Is this a new company, an existing company, or a service/agency team?
+3. Are you working solo on one machine, sharing privately with a team, or deploying publicly?
+4. Do you want full auto, hybrid, or tight manual control?
+
+Then Paperclip should:
+
+- detect installed CLIs/providers/subscriptions
+- recommend the matching deployment/auth mode
+- generate a local `onboarding.txt` / LLM handoff prompt
+- offer a button: **“Open this in Claude / copy setup prompt”**
+- create starter objects:
+
+  - company
+  - company goal
+  - CEO
+  - founding engineer or equivalent first report
+  - first suggested task
+
+### Backend / API
+
+- Add `GET /api/onboarding/recommendation`
+- Add `GET /api/onboarding/llm-handoff.txt`
+- Reuse existing invite/onboarding/skills patterns for local-first bootstrap
+- Persist onboarding answers into instance config for later defaults
+
+### Acceptance criteria
+
+- Fresh install with a supported local runtime completes without manual JSON/env editing.
+- User sees first live agent action before leaving onboarding.
+- A blank dashboard is no longer the default post-install state.
+- If a required dependency is missing, the error is prescriptive and fixable from the UI/CLI.
+
+### Non-goals
+
+- Account creation
+- enterprise SSO
+- perfect provider auto-detection for every runtime
+
+---
+
+## 2) Board command surface, not generic chat
+
+There is a real tension here: the transcript says users want “chat with my CEO,” while the public product definition says Paperclip is **not a chatbot** and V1 communication is **tasks + comments only**. At the same time, the repo is already exploring plugin infrastructure and even a chat plugin via plugin SSE streaming. The clean resolution is: **make the core surface conversational, but keep the data model task/thread-centric; reserve full chat as an optional plugin**. ([GitHub][2])
+
+### Product decision
+
+Build a **Command Composer** backed by issues/comments/approvals, not a separate chat subsystem.
+
+### What we want
+
+- “Talk to the CEO” feeling for the user.
+- Every conversation ends up attached to a real company object.
+- Strategy discussion can produce issues, artifacts, and approvals.
+
+### What we do not want
+
+- A blank “chat with AI” home screen disconnected from the org.
+- Yet another agent-chat product.
+
+### Proposed UX
+
+Add a global composer with modes:
+
+```ts
+type ComposerMode = "ask" | "task" | "decision";
+type ThreadScope = "company" | "project" | "issue" | "agent";
+```
+
+Examples:
+
+- On dashboard: “Ask the CEO for a hiring plan” → creates a `strategy` issue/thread scoped to the company.
+- On agent page: “Tell the designer to make this cleaner” → appends an instruction comment to an issue or spawns a new delegated task.
+- On approval page: “Why are you asking to hire?” → appends a board comment to the approval context.
+
+Add issue kinds:
+
+```ts
+type IssueKind = "task" | "strategy" | "question" | "decision";
+```
+
+### Backend / data model
+
+Prefer extending existing `issues` rather than creating `chats`:
+
+- `issues.kind`
+- `issues.scope`
+- optional `issues.target_agent_id`
+- comment metadata: `comment.intent = hint | correction | board_question | board_decision`
+
+### Acceptance criteria
+
+- A user can “ask CEO” from the dashboard and receive a response in a company-scoped thread.
+- From that thread, the user can create/approve tasks with one click.
+- No separate chat database is required for v1 of this feature.
+
+### Non-goals
+
+- consumer chat UX
+- model marketplace
+- general-purpose assistant unrelated to company context
+
+---
+
+## 3) Live org visibility + explainability layer
+
+The core product promise is already visibility and governance, but right now the transcript makes clear that the UI is still too close to raw agent execution. The repo already has org charts, activity, heartbeat runs, costs, and agent detail surfaces; the missing piece is the explanatory layer above them. ([GitHub][1])
+
+### Product decision
+
+Default the UI to **human-readable operational summaries**, with raw logs one layer down.
+
+### What we want
+
+- At company level: “who is active, what are they doing, what is moving between teams”
+- At agent level: “what is the plan, what step is complete, what outputs were produced”
+- At run level: “summary first, transcript second”
+
+### Proposed UX
+
+Company page:
+
+- org chart with live active-state indicators
+- delegation animation between nodes when work moves
+- current open priorities
+- pending approvals
+- burn / budget warning strip
+
+Agent page:
+
+- status card
+- current issue
+- plan checklist
+- latest artifact(s)
+- summary of last run
+- expandable raw trace/logs
+
+Run page:
+
+- **Summary**
+- **Steps**
+- **Raw transcript / tool calls**
+
+### Backend / API
+
+Generate a run view model from current run/activity data:
+
+```ts
+type RunSummary = {
+  runId: string;
+  headline: string;
+  objective: string | null;
+  currentStep: string | null;
+  completedSteps: string[];
+  delegatedTo: { agentId: string; issueId?: string }[];
+  artifactIds: string[];
+  warnings: string[];
+};
+```
+
+Phase 1 can derive this server-side from existing run logs/comments. Persist only if needed later.
+
+### Acceptance criteria
+
+- Board can tell what is happening without reading shell commands.
+- Raw logs are still accessible, but not the default surface.
+- First task / first hire / first completion moments are visibly celebrated.
+
+### Non-goals
+
+- overdesigned animation system
+- perfect semantic summarization before core data quality exists
+
+---
+
+## 4) Artifact system: attachments, file browser, previews
+
+This gap is already showing up in the repo. Storage is present, attachment endpoints exist, but current issues show that attachments are still effectively image-centric and comment attachment rendering is incomplete. At the same time, your transcript wants plans, docs, files, and generated web pages surfaced cleanly. ([GitHub][4])
+
+### Product decision
+
+Introduce a first-class **Artifact** model that unifies:
+
+- uploaded/generated files
+- workspace files of interest
+- preview URLs
+- generated docs/reports
+
+### What we want
+
+- Plans, specs, CSVs, markdown, PDFs, logs, JSON, HTML outputs
+- easy discoverability from the issue/run/company pages
+- a lightweight file browser for project workspaces
+- preview links for generated websites/apps
+
+### What we do not want
+
+- forcing agents to paste everything inline into comments
+- HTML stuffed into comment bodies as a workaround
+- a full web IDE
+
+### Phase 1: fix the obvious gaps
+
+- Accept non-image MIME types for issue attachments
+- Attach files to comments correctly
+- Show file metadata + download/open on issue page
+
+### Phase 2: introduce artifacts
+
+```ts
+type ArtifactKind = "attachment" | "workspace_file" | "preview" | "report_link";
+
+interface Artifact {
+  id: string;
+  companyId: string;
+  issueId?: string;
+  runId?: string;
+  agentId?: string;
+  kind: ArtifactKind;
+  title: string;
+  mimeType?: string;
+  filename?: string;
+  sizeBytes?: number;
+  storageKind: "local_disk" | "s3" | "external_url";
+  contentPath?: string;
+  previewUrl?: string;
+  metadata: Record<string, unknown>;
+}
+```
+
+### UX
+
+Issue page gets a **Deliverables** section:
+
+- Files
+- Reports
+- Preview links
+- Latest generated artifact highlighted at top
+
+Project page gets a **Files** tab:
+
+- folder tree
+- recent changes
+- “Open produced files” shortcut
+
+### Preview handling
+
+For HTML/static outputs:
+
+- local deploy → open local preview URL
+- shared/public deploy → host via configured preview service or static storage
+- preview URL is registered back onto the issue as an artifact
+
+### Acceptance criteria
+
+- Agents can attach `.md`, `.txt`, `.json`, `.csv`, `.pdf`, and `.html`.
+- Users can open/download them from the issue page.
+- A generated static site can be opened from an issue without hunting through the filesystem.
+
+### Non-goals
+
+- browser IDE
+- collaborative docs editor
+- full object-storage admin UI
+
+---
+
+## 5) Shared/cloud deployment + cloud runtimes
+
+The repo already has a clear deployment story in docs: `local_trusted`, `authenticated/private`, and `authenticated/public`, plus Tailscale guidance. The roadmap explicitly calls out cloud agents like Cursor / e2b. That means the next step is not inventing a deployment model; it is making the shared/cloud path canonical and production-usable. ([GitHub][5])
+
+### Product decision
+
+Make **shared/private deploy** and **public/cloud deploy** first-class supported modes, and add **remote runtime drivers** for cloud-executed agents.
+
+### What we want
+
+- one instance a team can actually share
+- local-first path that upgrades to private/public without a mental model change
+- remote agent execution for non-local runtimes
+
+### Proposed architecture
+
+Separate **control plane** from **execution runtime** more explicitly:
+
+```ts
+type RuntimeDriver = "local_process" | "remote_sandbox" | "webhook";
+
+interface ExecutionHandle {
+  externalRunId: string;
+  status: "queued" | "running" | "completed" | "failed" | "cancelled";
+  previewUrl?: string;
+  logsUrl?: string;
+}
+```
+
+First remote driver: `remote_sandbox` for e2b-style execution.
+
+### Deliverables
+
+- canonical deploy recipes:
+
+  - local solo
+  - shared private (Tailscale/private auth)
+  - public cloud (managed Postgres + object storage + public URL)
+
+- runtime health page
+- adapter/runtime capability matrix
+- one official reference deployment path
+
+### UX
+
+New “Deployment” settings page:
+
+- instance mode
+- auth/exposure
+- storage/database status
+- runtime drivers configured
+- health and reachability checks
+
+### Acceptance criteria
+
+- Two humans can log into one authenticated/private instance and use it concurrently.
+- A public deployment can run agents via at least one remote runtime.
+- `doctor` catches missing public/private config and gives concrete fixes.
+
+### Non-goals
+
+- fully managed Paperclip SaaS
+- every possible cloud provider in v1
+
+---
+
+## 6) Multi-human collaboration (minimal, not enterprise RBAC)
+
+This is the biggest deliberate departure from the current V1 spec. Publicly, V1 still says “single human board operator” and puts role-based human granularity out of scope. But the transcript is right that shared use is necessary if Paperclip is going to be real for teams. The key is to do a **minimal collaboration model**, not a giant permission system. ([GitHub][2])
+
+### Product decision
+
+Ship **coarse multi-user company memberships**, not fine-grained enterprise RBAC.
+
+### Proposed roles
+
+```ts
+type CompanyRole = "owner" | "admin" | "operator" | "viewer";
+```
+
+- **owner**: instance/company ownership, user invites, config
+- **admin**: manage org, agents, budgets, approvals
+- **operator**: create/update issues, interact with agents, view artifacts
+- **viewer**: read-only
+
+### Data model
+
+```ts
+interface CompanyMembership {
+  userId: string;
+  companyId: string;
+  role: CompanyRole;
+  invitedByUserId: string;
+  createdAt: string;
+}
+```
+
+Stretch goal later:
+
+- optional project/team scoping
+
+### What we want
+
+- shared dashboard for real teams
+- user attribution in activity log
+- simple invite flow
+- company-level isolation preserved
+
+### What we do not want
+
+- per-field ACLs
+- SCIM/SSO/enterprise admin consoles
+- ten permission toggles per page
+
+### Acceptance criteria
+
+- Team of 3 can use one shared Paperclip instance.
+- Every user action is attributed correctly in activity.
+- Company membership boundaries are enforced.
+- Viewer cannot mutate; operator/admin can.
+
+### Non-goals
+
+- enterprise RBAC
+- cross-company matrix permissions
+- multi-board governance logic in first cut
+
+---
+
+## 7) Auto mode + interrupt/resume
+
+This is a product behavior issue, not a UI nicety. If agents cannot keep working or accept course correction without restarting, the autonomy model feels fake.
+
+### Product decision
+
+Make auto mode and mid-run interruption first-class runtime semantics.
+
+### What we want
+
+- Auto mode that continues until blocked by approvals, budgets, or explicit pause.
+- Mid-run “you missed this” correction without losing session continuity.
+- Clear state when an agent is waiting, blocked, or paused.
+
+### Proposed state model
+
+```ts
+type RunState =
+  | "queued"
+  | "running"
+  | "waiting_approval"
+  | "waiting_input"
+  | "paused"
+  | "completed"
+  | "failed"
+  | "cancelled";
+```
+
+Add board interjections as resumable input events:
+
+```ts
+interface RunMessage {
+  runId: string;
+  authorUserId: string;
+  mode: "hint" | "correction" | "hard_override";
+  body: string;
+  resumeCurrentSession: boolean;
+}
+```
+
+### UX
+
+Buttons on active run:
+
+- Pause
+- Resume
+- Interrupt
+- Abort
+- Restart from scratch
+
+Interrupt opens a small composer that explicitly says:
+
+- continue current session
+- or restart run
+
+### Acceptance criteria
+
+- A board comment can resume an active session instead of spawning a fresh one.
+- Session ID remains stable for “continue” path.
+- UI clearly distinguishes blocked vs. waiting vs. paused.
+
+### Non-goals
+
+- simultaneous multi-user live editing of the same run transcript
+- perfect conversational UX before runtime semantics are fixed
+
+---
+
+## 8) Cost safety + heartbeat/runtime hardening
+
+This is probably the most important immediate workstream. The transcript says token burn is the highest pain, and the repo currently has active issues around budget enforcement evidence, onboarding/auth validation, and circuit-breaker style waste prevention. Public docs already promise hard budgets, and the issue tracker is pointing at the missing operational protections. ([GitHub][6])
+
+### Product decision
+
+Treat this as a **P0 runtime contract**, not a nice-to-have.
+
+### Part A: deterministic wake gating
+
+Do cheap, explicit work detection before invoking an LLM.
+
+```ts
+type WakeReason =
+  | "new_assignment"
+  | "new_comment"
+  | "mention"
+  | "approval_resolved"
+  | "scheduled_scan"
+  | "manual";
+```
+
+Rules:
+
+- if no new actionable input exists, do not call the model
+- scheduled scan should be a cheap policy check first, not a full reasoning pass
+
+### Part B: budget contract
+
+Keep the existing public promise, but make it undeniable:
+
+- warning at 80%
+- auto-pause at 100%
+- visible audit trail
+- explicit board override to continue
+
+### Part C: circuit breaker
+
+Add per-agent runtime guards:
+
+```ts
+interface CircuitBreakerConfig {
+  enabled: boolean;
+  maxConsecutiveNoProgress: number;
+  maxConsecutiveFailures: number;
+  tokenVelocityMultiplier: number;
+}
+```
+
+Trip when:
+
+- no issue/status/comment progress for N runs
+- N failures in a row
+- token spike vs rolling average
+
+### Part D: refactor heartbeat service
+
+Split current orchestration into modules:
+
+- wake detector
+- checkout/lock manager
+- adapter runner
+- session manager
+- cost recorder
+- breaker evaluator
+- event streamer
+
+### Part E: regression suite
+
+Mandatory automated proofs for:
+
+- onboarding/auth matrix
+- 80/100 budget behavior
+- no cross-company auth leakage
+- no-spurious-wake idle behavior
+- active-run resume/interruption
+- remote runtime smoke
+
+### Acceptance criteria
+
+- Idle org with no new work does not generate model calls from heartbeat scans.
+- 80% shows warning only.
+- 100% pauses the agent and blocks continued execution until override.
+- Circuit breaker pause is visible in audit/activity.
+- Runtime modules have explicit contracts and are testable independently.
+
+### Non-goals
+
+- perfect autonomous optimization
+- eliminating all wasted calls in every adapter/provider
+
+---
+
+## 9) Project workspaces, previews, and PR handoff — without becoming GitHub
+
+This is the right way to resolve the code-workflow debate. The repo already has worktree-local instances, project `workspaceStrategy.provisionCommand`, and an RFC for adapter-level git worktree isolation. That is the correct architectural direction: **project execution policies and workspace isolation**, not built-in PR review. ([GitHub][7])
+
+### Product decision
+
+Paperclip should manage the **issue → workspace → preview/PR → review handoff** lifecycle, but leave diffs/review/merge to external tools.
+
+### Proposed config
+
+Prefer repo-local project config:
+
+```yaml
+# .paperclip/project.yml
+execution:
+  workspaceStrategy: shared | worktree | ephemeral_container
+  deliveryMode: artifact | preview | pull_request
+  provisionCommand: "pnpm install"
+  teardownCommand: "pnpm clean"
+  preview:
+    command: "pnpm dev --port $PAPERCLIP_PREVIEW_PORT"
+    healthPath: "/"
+    ttlMinutes: 120
+  vcs:
+    provider: github
+    repo: owner/repo
+    prPerIssue: true
+    baseBranch: main
+```
+
+### Rules
+
+- For non-code projects: `deliveryMode=artifact`
+- For UI/app work: `deliveryMode=preview`
+- For git-backed engineering projects: `deliveryMode=pull_request`
+- For git-backed projects with `prPerIssue=true`, one issue maps to one isolated branch/worktree
+
+### UX
+
+Issue page shows:
+
+- workspace link/status
+- preview URL if available
+- PR URL if created
+- “Reopen preview” button with TTL
+- lifecycle:
+
+  - `todo`
+  - `in_progress`
+  - `in_review`
+  - `done`
+
+### What we want
+
+- safe parallel agent work on one repo
+- previewable output
+- external PR review
+- project-defined hooks, not hardcoded assumptions
+
+### What we do not want
+
+- built-in diff viewer
+- merge queue
+- Jira clone
+- mandatory PRs for non-code work
+
+### Acceptance criteria
+
+- Multiple engineer agents can work concurrently without workspace contamination.
+- When a project is in PR mode, the issue contains branch/worktree/preview/PR metadata.
+- Preview can be reopened on demand until TTL expires.
+
+### Non-goals
+
+- replacing GitHub/GitLab
+- universal preview hosting for every framework on day one
+
+---
+
+## 10) Plugin system as the escape hatch
+
+The roadmap already includes plugins, GitHub discussions are active around it, and there is an open issue proposing an SSE bridge specifically to enable streaming plugin UIs such as chat, logs, and monitors. This is exactly the right place for optional surfaces. ([GitHub][1])
+
+### Product decision
+
+Keep the control-plane core thin; put optional high-variance experiences into plugins.
+
+### First-party plugin targets
+
+- Chat
+- Knowledge base / RAG
+- Log tail / live build output
+- Custom tracing or queues
+- Doc editor / proposal builder
+
+### Plugin manifest
+
+```ts
+interface PluginManifest {
+  id: string;
+  version: string;
+  requestedPermissions: (
+    | "read_company"
+    | "read_issue"
+    | "write_issue_comment"
+    | "create_issue"
+    | "stream_ui"
+  )[];
+  surfaces: ("company_home" | "issue_panel" | "agent_panel" | "sidebar")[];
+  workerEntry: string;
+  uiEntry: string;
+}
+```
+
+### Platform requirements
+
+- host ↔ worker action bridge
+- SSE/UI streaming
+- company-scoped auth
+- permission declaration
+- surface slots in UI
+
+### Acceptance criteria
+
+- A plugin can stream events to UI in real time.
+- A chat plugin can converse without requiring chat to become the core Paperclip product.
+- Plugin permissions are company-scoped and auditable.
+
+### Non-goals
+
+- plugins mutating core schema directly
+- arbitrary privileged code execution without explicit permissions
+
+---
+
+## Priority order I would use
+
+Given the repo state and the transcript, I would sequence it like this:
+
+**P0**
+
+1. Cost safety + heartbeat hardening
+2. Guided onboarding + first-job magic
+3. Shared/cloud deployment foundation
+4. Artifact phase 1: non-image attachments + deliverables surfacing
+
+**P1** 5. Board command surface 6. Visibility/explainability layer 7. Auto mode + interrupt/resume 8. Minimal multi-user collaboration
+
+**P2** 9. Project workspace / preview / PR lifecycle 10. Plugin system + optional chat plugin 11. Template/preset expansion for startup vs agency vs internal-team onboarding
+
+Why this order: the current repo is already getting pressure on onboarding failures, auth/onboarding validation, budget enforcement, and wasted token burn. If those are shaky, everything else feels impressive but unsafe. ([GitHub][3])
+
+## Bottom line
+
+The best synthesis is:
+
+- **Keep** Paperclip as the board-level control plane.
+- **Do not** make chat, code review, or workflow-building the core identity.
+- **Do** make the product feel conversational, visible, output-oriented, and shared.
+- **Do** make coding workflows an integration surface via workspaces/previews/PR links.
+- **Use plugins** for richer edges like chat and knowledge.
+
+That keeps the repo’s current product direction intact while solving almost every pain surfaced in the transcript.
+
+### Key references
+
+- README / positioning / roadmap / product boundaries. ([GitHub][1])
+- Product definition. ([GitHub][8])
+- V1 implementation spec and explicit non-goals. ([GitHub][2])
+- Core concepts and architecture. ([GitHub][9])
+- Deployment modes / Tailscale / local-to-cloud path. ([GitHub][5])
+- Developing guide: worktree-local instances, provision hooks, onboarding endpoints. ([GitHub][7])
+- Current issue pressure: onboarding failure, auth/onboarding validation, budget enforcement, circuit breaker, attachment gaps, plugin chat. ([GitHub][3])
+
+[1]: https://github.com/paperclipai/paperclip "https://github.com/paperclipai/paperclip"
+[2]: https://github.com/paperclipai/paperclip/blob/master/doc/SPEC-implementation.md "https://github.com/paperclipai/paperclip/blob/master/doc/SPEC-implementation.md"
+[3]: https://github.com/paperclipai/paperclip/issues/704 "https://github.com/paperclipai/paperclip/issues/704"
+[4]: https://github.com/paperclipai/paperclip/blob/master/docs/deploy/tailscale-private-access.md "https://github.com/paperclipai/paperclip/blob/master/docs/deploy/tailscale-private-access.md"
+[5]: https://github.com/paperclipai/paperclip/blob/master/docs/deploy/deployment-modes.md "https://github.com/paperclipai/paperclip/blob/master/docs/deploy/deployment-modes.md"
+[6]: https://github.com/paperclipai/paperclip/issues/692 "https://github.com/paperclipai/paperclip/issues/692"
+[7]: https://github.com/paperclipai/paperclip/blob/master/doc/DEVELOPING.md "https://github.com/paperclipai/paperclip/blob/master/doc/DEVELOPING.md"
+[8]: https://github.com/paperclipai/paperclip/blob/master/doc/PRODUCT.md "https://github.com/paperclipai/paperclip/blob/master/doc/PRODUCT.md"
+[9]: https://github.com/paperclipai/paperclip/blob/master/docs/start/core-concepts.md "https://github.com/paperclipai/paperclip/blob/master/docs/start/core-concepts.md"

doc/plans/2026-03-13-paperclip-skill-tightening-plan.md

@@ -0,0 +1,186 @@
+# Paperclip Skill Tightening Plan
+
+## Status
+
+Deferred follow-up. Do not include in the current token-optimization PR beyond documenting the plan.
+
+## Why This Is Deferred
+
+The `paperclip` skill is part of the critical control-plane safety surface. Tightening it may reduce fresh-session token use, but it also carries prompt-regression risk. We do not yet have evals that would let us safely prove behavior preservation across assignment handling, checkout rules, comment etiquette, approval workflows, and escalation paths.
+
+The current PR should ship the lower-risk infrastructure wins first:
+
+- telemetry normalization
+- safe session reuse
+- incremental issue/comment context
+- bootstrap versus heartbeat prompt separation
+- Codex worktree isolation
+
+## Current Problem
+
+Fresh runs still spend substantial input tokens even after the context-path fixes. The remaining large startup cost appears to come from loading the full `paperclip` skill and related instruction surface into context at run start.
+
+The skill currently mixes three kinds of content in one file:
+
+- hot-path heartbeat procedure used on nearly every run
+- critical policy and safety invariants
+- rare workflow/reference material that most runs do not need
+
+That structure is safe but expensive.
+
+## Goals
+
+- reduce first-run instruction tokens without weakening agent safety
+- preserve all current Paperclip control-plane capabilities
+- keep common heartbeat behavior explicit and easy for agents to follow
+- move rare workflows and reference material out of the hot path
+- create a structure that can later be evaluated systematically
+
+## Non-Goals
+
+- changing Paperclip API semantics
+- removing required governance rules
+- deleting rare workflows
+- changing agent defaults in the current PR
+
+## Recommended Direction
+
+### 1. Split Hot Path From Lookup Material
+
+Restructure the skill into:
+
+- an always-loaded core section for the common heartbeat loop
+- on-demand material for infrequent workflows and deep reference
+
+The core should cover only what is needed on nearly every wake:
+
+- auth and required headers
+- inbox-first assignment retrieval
+- mandatory checkout behavior
+- `heartbeat-context` first
+- incremental comment retrieval rules
+- mention/self-assign exception
+- blocked-task dedup
+- status/comment/release expectations before exit
+
+### 2. Normalize The Skill Around One Canonical Procedure
+
+The same rules are currently expressed multiple times across:
+
+- heartbeat steps
+- critical rules
+- endpoint reference
+- workflow examples
+
+Refactor so each operational fact has one primary home:
+
+- procedure
+- invariant list
+- appendix/reference
+
+This reduces prompt weight and lowers the chance of internal instruction drift.
+
+### 3. Compress Prose Into High-Signal Instruction Forms
+
+Rewrite the hot path using compact operational forms:
+
+- short ordered checklist
+- flat invariant list
+- minimal examples only where ambiguity would be risky
+
+Reduce:
+
+- narrative explanation
+- repeated warnings already covered elsewhere
+- large example payloads for common operations
+- long endpoint matrices in the main body
+
+### 4. Move Rare Workflows Behind Explicit Triggers
+
+These workflows should remain available but should not dominate fresh-run context:
+
+- OpenClaw invite flow
+- project setup flow
+- planning `<plan/>` writeback flow
+- instructions-path update flow
+- detailed link-formatting examples
+
+Recommended approach:
+
+- keep a short pointer in the main skill
+- move detailed procedures into sibling skills or referenced docs that agents read only when needed
+
+### 5. Separate Policy From Reference
+
+The skill should distinguish:
+
+- mandatory operating rules
+- endpoint lookup/reference
+- business-process playbooks
+
+That separation makes it easier to evaluate prompt changes later and lets adapters or orchestration choose what must always be loaded.
+
+## Proposed Target Structure
+
+1. Purpose and authentication
+2. Compact heartbeat procedure
+3. Hard invariants
+4. Required comment/update style
+5. Triggered workflow index
+6. Appendix/reference
+
+## Rollout Plan
+
+### Phase 1. Inventory And Measure
+
+- annotate the current skill by section and estimate token weight
+- identify which sections are truly hot-path versus rare
+- capture representative runs to compare before/after prompt size and behavior
+
+### Phase 2. Structural Refactor Without Semantic Changes
+
+- rewrite the main skill into the target structure
+- preserve all existing rules and capabilities
+- move rare workflow details into referenced companion material
+- keep wording changes conservative
+
+### Phase 3. Validate Against Real Scenarios
+
+Run scenario checks for:
+
+- normal assigned heartbeat
+- comment-triggered wake
+- blocked-task dedup behavior
+- approval-resolution wake
+- delegation/subtask creation
+- board handoff back to user
+- plan-request handling
+
+### Phase 4. Decide Default Loading Strategy
+
+After validation, decide whether:
+
+- the entire main skill still loads by default, or
+- only the compact core loads by default and rare sections are fetched on demand
+
+Do not change this loading policy without validation.
+
+## Risks
+
+- prompt degradation on control-plane safety rules
+- agents forgetting rare but important workflows
+- accidental removal of repeated wording that was carrying useful behavior
+- introducing ambiguous instruction precedence between the core skill and companion materials
+
+## Preconditions Before Implementation
+
+- define acceptance scenarios for control-plane correctness
+- add at least lightweight eval or scripted scenario coverage for key Paperclip flows
+- confirm how adapter/bootstrap layering should load skill content versus references
+
+## Success Criteria
+
+- materially lower first-run input tokens for Paperclip-coordinated agents
+- no regression in checkout discipline, issue updates, blocked handling, or delegation
+- no increase in malformed API usage or ownership mistakes
+- agents still complete rare workflows correctly when explicitly asked

doc/plans/2026-03-13-plugin-kitchen-sink-example.md

@@ -0,0 +1,699 @@
+# Kitchen Sink Plugin Plan
+
+## Goal
+
+Add a new first-party example plugin, `Kitchen Sink (Example)`, that demonstrates every currently implemented Paperclip plugin API surface in one place.
+
+This plugin is meant to be:
+
+- a living reference implementation for contributors
+- a manual test harness for the plugin runtime
+- a discoverable demo of what plugins can actually do today
+
+It is not meant to be a polished end-user product plugin.
+
+## Why
+
+The current plugin system has a real API surface, but it is spread across:
+
+- SDK docs
+- SDK types
+- plugin spec prose
+- two example plugins that each show only a narrow slice
+
+That makes it hard to answer basic questions like:
+
+- what can plugins render?
+- what can plugin workers actually do?
+- which surfaces are real versus aspirational?
+- how should a new plugin be structured in this repo?
+
+The kitchen-sink plugin should answer those questions by example.
+
+## Success Criteria
+
+The plugin is successful if a contributor can install it and, without reading the SDK first, discover and exercise the current plugin runtime surface area from inside Paperclip.
+
+Concretely:
+
+- it installs from the bundled examples list
+- it exposes at least one demo for every implemented worker API surface
+- it exposes at least one demo for every host-mounted UI surface
+- it clearly labels local-only / trusted-only demos
+- it is safe enough for local development by default
+- it doubles as a regression harness for plugin runtime changes
+
+## Constraints
+
+- Keep it instance-installed, not company-installed.
+- Treat this as a trusted/local example plugin.
+- Do not rely on cloud-safe runtime assumptions.
+- Avoid destructive defaults.
+- Avoid irreversible mutations unless they are clearly labeled and easy to undo.
+
+## Source Of Truth For This Plan
+
+This plan is based on the currently implemented SDK/types/runtime, not only the long-horizon spec.
+
+Primary references:
+
+- `packages/plugins/sdk/README.md`
+- `packages/plugins/sdk/src/types.ts`
+- `packages/plugins/sdk/src/ui/types.ts`
+- `packages/shared/src/constants.ts`
+- `packages/shared/src/types/plugin.ts`
+
+## Current Surface Inventory
+
+### Worker/runtime APIs to demonstrate
+
+These are the concrete `ctx` clients currently exposed by the SDK:
+
+- `ctx.config`
+- `ctx.events`
+- `ctx.jobs`
+- `ctx.launchers`
+- `ctx.http`
+- `ctx.secrets`
+- `ctx.assets`
+- `ctx.activity`
+- `ctx.state`
+- `ctx.entities`
+- `ctx.projects`
+- `ctx.companies`
+- `ctx.issues`
+- `ctx.agents`
+- `ctx.goals`
+- `ctx.data`
+- `ctx.actions`
+- `ctx.streams`
+- `ctx.tools`
+- `ctx.metrics`
+- `ctx.logger`
+
+### UI surfaces to demonstrate
+
+Surfaces defined in the SDK:
+
+- `page`
+- `settingsPage`
+- `dashboardWidget`
+- `sidebar`
+- `sidebarPanel`
+- `detailTab`
+- `taskDetailView`
+- `projectSidebarItem`
+- `toolbarButton`
+- `contextMenuItem`
+- `commentAnnotation`
+- `commentContextMenuItem`
+
+### Current host confidence
+
+Confirmed or strongly indicated as mounted in the current app:
+
+- `page`
+- `settingsPage`
+- `dashboardWidget`
+- `detailTab`
+- `projectSidebarItem`
+- comment surfaces
+- launcher infrastructure
+
+Need explicit validation before claiming full demo coverage:
+
+- `sidebar`
+- `sidebarPanel`
+- `taskDetailView`
+- `toolbarButton` as direct slot, distinct from launcher placement
+- `contextMenuItem` as direct slot, distinct from comment menu and launcher placement
+
+The implementation should keep a small validation checklist for these before we call the plugin "complete".
+
+## Plugin Concept
+
+The plugin should be named:
+
+- display name: `Kitchen Sink (Example)`
+- package: `@paperclipai/plugin-kitchen-sink-example`
+- plugin id: `paperclip.kitchen-sink-example` or `paperclip-kitchen-sink-example`
+
+Recommendation: use `paperclip-kitchen-sink-example` to match current in-repo example naming style.
+
+Category mix:
+
+- `ui`
+- `automation`
+- `workspace`
+- `connector`
+
+That is intentionally broad because the point is coverage.
+
+## UX Shape
+
+The plugin should have one main full-page demo console plus smaller satellites on other surfaces.
+
+### 1. Plugin page
+
+Primary route: the plugin `page` surface should be the central dashboard for all demos.
+
+Recommended page sections:
+
+- `Overview`
+  - what this plugin demonstrates
+  - current capabilities granted
+  - current host context
+- `UI Surfaces`
+  - links explaining where each other surface should appear
+- `Data + Actions`
+  - buttons and forms for bridge-driven worker demos
+- `Events + Streams`
+  - emit event
+  - watch event log
+  - stream demo output
+- `Paperclip Domain APIs`
+  - companies
+  - projects/workspaces
+  - issues
+  - goals
+  - agents
+- `Local Workspace + Process`
+  - file listing
+  - file read/write scratch area
+  - child process demo
+- `Jobs + Webhooks + Tools`
+  - job status
+  - webhook URL and recent deliveries
+  - declared tools
+- `State + Entities + Assets`
+  - scoped state editor
+  - plugin entity inspector
+  - upload/generated asset demo
+- `Observability`
+  - metrics written
+  - activity log samples
+  - latest worker logs
+
+### 2. Dashboard widget
+
+A compact widget on the main dashboard should show:
+
+- plugin health
+- count of demos exercised
+- recent event/stream activity
+- shortcut to the full plugin page
+
+### 3. Project sidebar item
+
+Add a `Kitchen Sink` link under each project that deep-links into a project-scoped plugin tab.
+
+### 4. Detail tabs
+
+Use detail tabs to demonstrate entity-context rendering on:
+
+- `project`
+- `issue`
+- `agent`
+- `goal`
+
+Each tab should show:
+
+- the host context it received
+- the relevant entity fetch via worker bridge
+- one small action scoped to that entity
+
+### 5. Comment surfaces
+
+Use issue comment demos to prove comment-specific extension points:
+
+- `commentAnnotation`
+  - render parsed metadata below each comment
+  - show comment id, issue id, and a small derived status
+- `commentContextMenuItem`
+  - add a menu action like `Copy Context To Kitchen Sink`
+  - action writes a plugin entity or state record for later inspection
+
+### 6. Settings page
+
+Custom `settingsPage` should be intentionally simple and operational:
+
+- `About`
+- `Danger / Trust Model`
+- demo toggles
+- local process defaults
+- workspace scratch-path behavior
+- secret reference inputs
+- event/job/webhook sample config
+
+This plugin should also keep the generic plugin settings `Status` tab useful by writing health, logs, and metrics.
+
+## Feature Matrix
+
+Each implemented worker API should have a visible demo.
+
+### `ctx.config`
+
+Demo:
+
+- read live config
+- show config JSON
+- react to config changes without restart where possible
+
+### `ctx.events`
+
+Demos:
+
+- emit a plugin event
+- subscribe to plugin events
+- subscribe to a core Paperclip event such as `issue.created`
+- show recent received events in a timeline
+
+### `ctx.jobs`
+
+Demos:
+
+- one scheduled heartbeat-style demo job
+- one manual run button from the UI if host supports manual job trigger
+- show last run result and timestamps
+
+### `ctx.launchers`
+
+Demos:
+
+- declare launchers in manifest
+- optionally register one runtime launcher from the worker
+- show launcher metadata on the plugin page
+
+### `ctx.http`
+
+Demo:
+
+- make a simple outbound GET request to a safe endpoint
+- show status code, latency, and JSON result
+
+Recommendation: default to a Paperclip-local endpoint or a stable public echo endpoint to avoid flaky docs.
+
+### `ctx.secrets`
+
+Demo:
+
+- operator enters a secret reference in config
+- plugin resolves it on demand
+- UI only shows masked result length / success status, never raw secret
+
+### `ctx.assets`
+
+Demos:
+
+- generate a text asset from the UI
+- optionally upload a tiny JSON blob or screenshot-like text file
+- show returned asset URL
+
+### `ctx.activity`
+
+Demo:
+
+- button to write a plugin activity log entry against current company/entity
+
+### `ctx.state`
+
+Demos:
+
+- instance-scoped state
+- company-scoped state
+- project-scoped state
+- issue-scoped state
+- delete/reset controls
+
+Use a small state inspector/editor on the plugin page.
+
+### `ctx.entities`
+
+Demos:
+
+- create plugin-owned sample records
+- list/filter them
+- show one realistic use case such as "copied comments" or "demo sync records"
+
+### `ctx.projects`
+
+Demos:
+
+- list projects
+- list project workspaces
+- resolve primary workspace
+- resolve workspace for issue
+
+### `ctx.companies`
+
+Demo:
+
+- list companies and show current selected company
+
+### `ctx.issues`
+
+Demos:
+
+- list issues in current company
+- create issue
+- update issue status/title
+- list comments
+- create comment
+
+### `ctx.agents`
+
+Demos:
+
+- list agents
+- invoke one agent with a test prompt
+- pause/resume where safe
+
+Agent mutation controls should be behind an explicit warning.
+
+### `ctx.agents.sessions`
+
+Demos:
+
+- create agent chat session
+- send message
+- stream events back to the UI
+- close session
+
+This is a strong candidate for the best "wow" demo on the plugin page.
+
+### `ctx.goals`
+
+Demos:
+
+- list goals
+- create goal
+- update status/title
+
+### `ctx.data`
+
+Use throughout the plugin for all read-side bridge demos.
+
+### `ctx.actions`
+
+Use throughout the plugin for all mutation-side bridge demos.
+
+### `ctx.streams`
+
+Demos:
+
+- live event log stream
+- token-style stream from an agent session relay
+- fake progress stream for a long-running action
+
+### `ctx.tools`
+
+Demos:
+
+- declare 2-3 simple agent tools
+- tool 1: echo/diagnostics
+- tool 2: project/workspace summary
+- tool 3: create issue or write plugin state
+
+The plugin page should list declared tools and show example input payloads.
+
+### `ctx.metrics`
+
+Demo:
+
+- write a sample metric on each major demo action
+- surface a small recent metrics table in the plugin page
+
+### `ctx.logger`
+
+Demo:
+
+- every action logs structured entries
+- plugin settings `Status` page then doubles as the log viewer
+
+## Local Workspace And Process Demos
+
+The plugin SDK intentionally leaves file/process operations to the plugin itself once it has workspace metadata.
+
+The kitchen-sink plugin should demonstrate that explicitly.
+
+### Workspace demos
+
+- list files from a selected workspace
+- read a file
+- write to a plugin-owned scratch file
+- optionally search files with `rg` if available
+
+### Process demos
+
+- run a short-lived command like `pwd`, `ls`, or `git status`
+- stream stdout/stderr back to UI
+- show exit code and timing
+
+Important safeguards:
+
+- default commands must be read-only
+- no shell interpolation from arbitrary free-form input in v1
+- provide a curated command list or a strongly validated command form
+- clearly label this area as local-only and trusted-only
+
+## Proposed Manifest Coverage
+
+The plugin should aim to declare:
+
+- `page`
+- `settingsPage`
+- `dashboardWidget`
+- `detailTab` for `project`, `issue`, `agent`, `goal`
+- `projectSidebarItem`
+- `commentAnnotation`
+- `commentContextMenuItem`
+
+Then, after host validation, add if supported:
+
+- `sidebar`
+- `sidebarPanel`
+- `taskDetailView`
+- `toolbarButton`
+- `contextMenuItem`
+
+It should also declare one or more `ui.launchers` entries to exercise launcher behavior independently of slot rendering.
+
+## Proposed Package Layout
+
+New package:
+
+- `packages/plugins/examples/plugin-kitchen-sink-example/`
+
+Expected files:
+
+- `package.json`
+- `README.md`
+- `tsconfig.json`
+- `src/index.ts`
+- `src/manifest.ts`
+- `src/worker.ts`
+- `src/ui/index.tsx`
+- `src/ui/components/...`
+- `src/ui/hooks/...`
+- `src/lib/...`
+- optional `scripts/build-ui.mjs` if UI bundling needs esbuild
+
+## Proposed Internal Architecture
+
+### Worker modules
+
+Recommended split:
+
+- `src/worker.ts`
+  - plugin definition and wiring
+- `src/worker/data.ts`
+  - `ctx.data.register(...)`
+- `src/worker/actions.ts`
+  - `ctx.actions.register(...)`
+- `src/worker/events.ts`
+  - event subscriptions and event log buffer
+- `src/worker/jobs.ts`
+  - scheduled job handlers
+- `src/worker/tools.ts`
+  - tool declarations and handlers
+- `src/worker/local-runtime.ts`
+  - file/process demos
+- `src/worker/demo-store.ts`
+  - helpers for state/entities/assets/metrics
+
+### UI modules
+
+Recommended split:
+
+- `src/ui/index.tsx`
+  - exported slot components
+- `src/ui/page/KitchenSinkPage.tsx`
+- `src/ui/settings/KitchenSinkSettingsPage.tsx`
+- `src/ui/widgets/KitchenSinkDashboardWidget.tsx`
+- `src/ui/tabs/ProjectKitchenSinkTab.tsx`
+- `src/ui/tabs/IssueKitchenSinkTab.tsx`
+- `src/ui/tabs/AgentKitchenSinkTab.tsx`
+- `src/ui/tabs/GoalKitchenSinkTab.tsx`
+- `src/ui/comments/KitchenSinkCommentAnnotation.tsx`
+- `src/ui/comments/KitchenSinkCommentMenuItem.tsx`
+- `src/ui/shared/...`
+
+## Configuration Schema
+
+The plugin should have a substantial but understandable `instanceConfigSchema`.
+
+Recommended config fields:
+
+- `enableDangerousDemos`
+- `enableWorkspaceDemos`
+- `enableProcessDemos`
+- `showSidebarEntry`
+- `showSidebarPanel`
+- `showProjectSidebarItem`
+- `showCommentAnnotation`
+- `showCommentContextMenuItem`
+- `showToolbarLauncher`
+- `defaultDemoCompanyId` optional
+- `secretRefExample`
+- `httpDemoUrl`
+- `processAllowedCommands`
+- `workspaceScratchSubdir`
+
+Defaults should keep risky behavior off.
+
+## Safety Defaults
+
+Default posture:
+
+- UI and read-only demos on
+- mutating domain demos on but explicitly labeled
+- process demos off by default
+- no arbitrary shell input by default
+- no raw secret rendering ever
+
+## Phased Build Plan
+
+### Phase 1: Core plugin skeleton
+
+- scaffold package
+- add manifest, worker, UI entrypoints
+- add README
+- make it appear in bundled examples list
+
+### Phase 2: Core, confirmed UI surfaces
+
+- plugin page
+- settings page
+- dashboard widget
+- project sidebar item
+- detail tabs
+
+### Phase 3: Core worker APIs
+
+- config
+- state
+- entities
+- companies/projects/issues/goals
+- data/actions
+- metrics/logger/activity
+
+### Phase 4: Real-time and automation APIs
+
+- streams
+- events
+- jobs
+- webhooks
+- agent sessions
+- tools
+
+### Phase 5: Local trusted runtime demos
+
+- workspace file demos
+- child process demos
+- guarded by config
+
+### Phase 6: Secondary UI surfaces
+
+- comment annotation
+- comment context menu item
+- launchers
+
+### Phase 7: Validation-only surfaces
+
+Validate whether the current host truly mounts:
+
+- `sidebar`
+- `sidebarPanel`
+- `taskDetailView`
+- direct-slot `toolbarButton`
+- direct-slot `contextMenuItem`
+
+If mounted, add demos.
+If not mounted, document them as SDK-defined but host-pending.
+
+## Documentation Deliverables
+
+The plugin should ship with a README that includes:
+
+- what it demonstrates
+- which surfaces are local-only
+- how to install it
+- where each UI surface should appear
+- a mapping from demo card to SDK API
+
+It should also be referenced from plugin docs as the "reference everything plugin".
+
+## Testing And Verification
+
+Minimum verification:
+
+- package typecheck/build
+- install from bundled example list
+- page loads
+- widget appears
+- project tab appears
+- comment surfaces render
+- settings page loads
+- key actions succeed
+
+Recommended manual checklist:
+
+- create issue from plugin
+- create goal from plugin
+- emit and receive plugin event
+- stream action output
+- open agent session and receive streamed reply
+- upload an asset
+- write plugin activity log
+- run a safe local process demo
+
+## Open Questions
+
+1. Should the process demo remain curated-command-only in the first pass?
+   Recommendation: yes.
+
+2. Should the plugin create throwaway "kitchen sink demo" issues/goals automatically?
+   Recommendation: no. Make creation explicit.
+
+3. Should we expose unsupported-but-typed surfaces in the UI even if host mounting is not wired?
+   Recommendation: yes, but label them as `SDK-defined / host validation pending`.
+
+4. Should agent mutation demos include pause/resume by default?
+   Recommendation: probably yes, but behind a warning block.
+
+5. Should this plugin be treated as a supported regression harness in CI later?
+   Recommendation: yes. Long term, this should be the plugin-runtime smoke test package.
+
+## Recommended Next Step
+
+If this plan looks right, the next implementation pass should start by building only:
+
+- package skeleton
+- page
+- settings page
+- dashboard widget
+- one project detail tab
+- one issue detail tab
+- the basic worker/action/data/state/event scaffolding
+
+That is enough to lock the architecture before filling in every demo surface.

doc/plans/2026-03-14-adapter-skill-sync-rollout.md

@@ -0,0 +1,399 @@
+# 2026-03-14 Adapter Skill Sync Rollout
+
+Status: Proposed
+Date: 2026-03-14
+Audience: Product and engineering
+Related:
+- `doc/plans/2026-03-14-skills-ui-product-plan.md`
+- `doc/plans/2026-03-13-company-import-export-v2.md`
+- `docs/companies/companies-spec.md`
+
+## 1. Purpose
+
+This document defines the rollout plan for adapter-wide skill support in Paperclip.
+
+The goal is not just “show a skills tab.” The goal is:
+
+- every adapter has a deliberate skill-sync truth model
+- the UI tells the truth for that adapter
+- Paperclip stores desired skill state consistently even when the adapter cannot fully reconcile it
+- unsupported adapters degrade clearly and safely
+
+## 2. Current Adapter Matrix
+
+Paperclip currently has these adapters:
+
+- `claude_local`
+- `codex_local`
+- `cursor_local`
+- `gemini_local`
+- `opencode_local`
+- `pi_local`
+- `openclaw_gateway`
+
+The current skill API supports:
+
+- `unsupported`
+- `persistent`
+- `ephemeral`
+
+Current implementation state:
+
+- `codex_local`: implemented, `persistent`
+- `claude_local`: implemented, `ephemeral`
+- `cursor_local`: not yet implemented, but technically suited to `persistent`
+- `gemini_local`: not yet implemented, but technically suited to `persistent`
+- `pi_local`: not yet implemented, but technically suited to `persistent`
+- `opencode_local`: not yet implemented; likely `persistent`, but with special handling because it currently injects into Claude’s shared skills home
+- `openclaw_gateway`: not yet implemented; blocked on gateway protocol support, so `unsupported` for now
+
+## 3. Product Principles
+
+1. Desired skills live in Paperclip for every adapter.
+2. Adapters may expose different truth models, and the UI must reflect that honestly.
+3. Persistent adapters should read and reconcile actual installed state.
+4. Ephemeral adapters should report effective runtime state, not pretend they own a persistent install.
+5. Shared-home adapters need stronger safeguards than isolated-home adapters.
+6. Gateway or cloud adapters must not fake local filesystem sync.
+
+## 4. Adapter Classification
+
+### 4.1 Persistent local-home adapters
+
+These adapters have a stable local skills directory that Paperclip can read and manage.
+
+Candidates:
+
+- `codex_local`
+- `cursor_local`
+- `gemini_local`
+- `pi_local`
+- `opencode_local` with caveats
+
+Expected UX:
+
+- show actual installed skills
+- show managed vs external skills
+- support `sync`
+- support stale removal
+- preserve unknown external skills
+
+### 4.2 Ephemeral mount adapters
+
+These adapters do not have a meaningful Paperclip-owned persistent install state.
+
+Current adapter:
+
+- `claude_local`
+
+Expected UX:
+
+- show desired Paperclip skills
+- show any discoverable external dirs if available
+- say “mounted on next run” instead of “installed”
+- do not imply a persistent adapter-owned install state
+
+### 4.3 Unsupported / remote adapters
+
+These adapters cannot support skill sync without new external capabilities.
+
+Current adapter:
+
+- `openclaw_gateway`
+
+Expected UX:
+
+- company skill library still works
+- agent attachment UI still works at the desired-state level
+- actual adapter state is `unsupported`
+- sync button is disabled or replaced with explanatory text
+
+## 5. Per-Adapter Plan
+
+### 5.1 Codex Local
+
+Target mode:
+
+- `persistent`
+
+Current state:
+
+- already implemented
+
+Requirements to finish:
+
+- keep as reference implementation
+- tighten tests around external custom skills and stale removal
+- ensure imported company skills can be attached and synced without manual path work
+
+Success criteria:
+
+- list installed managed and external skills
+- sync desired skills into `CODEX_HOME/skills`
+- preserve external user-managed skills
+
+### 5.2 Claude Local
+
+Target mode:
+
+- `ephemeral`
+
+Current state:
+
+- already implemented
+
+Requirements to finish:
+
+- polish status language in UI
+- clearly distinguish “desired” from “mounted on next run”
+- optionally surface configured external skill dirs if Claude exposes them
+
+Success criteria:
+
+- desired skills stored in Paperclip
+- selected skills mounted per run
+- no misleading “installed” language
+
+### 5.3 Cursor Local
+
+Target mode:
+
+- `persistent`
+
+Technical basis:
+
+- runtime already injects Paperclip skills into `~/.cursor/skills`
+
+Implementation work:
+
+1. Add `listSkills` for Cursor.
+2. Add `syncSkills` for Cursor.
+3. Reuse the same managed-symlink pattern as Codex.
+4. Distinguish:
+   - managed Paperclip skills
+   - external skills already present
+   - missing desired skills
+   - stale managed skills
+
+Testing:
+
+- unit tests for discovery
+- unit tests for sync and stale removal
+- verify shared auth/session setup is not disturbed
+
+Success criteria:
+
+- Cursor agents show real installed state
+- syncing from the agent Skills tab works
+
+### 5.4 Gemini Local
+
+Target mode:
+
+- `persistent`
+
+Technical basis:
+
+- runtime already injects Paperclip skills into `~/.gemini/skills`
+
+Implementation work:
+
+1. Add `listSkills` for Gemini.
+2. Add `syncSkills` for Gemini.
+3. Reuse managed-symlink conventions from Codex/Cursor.
+4. Verify auth remains untouched while skills are reconciled.
+
+Potential caveat:
+
+- if Gemini treats that skills directory as shared user state, the UI should warn before removing stale managed skills
+
+Success criteria:
+
+- Gemini agents can reconcile desired vs actual skill state
+
+### 5.5 Pi Local
+
+Target mode:
+
+- `persistent`
+
+Technical basis:
+
+- runtime already injects Paperclip skills into `~/.pi/agent/skills`
+
+Implementation work:
+
+1. Add `listSkills` for Pi.
+2. Add `syncSkills` for Pi.
+3. Reuse managed-symlink helpers.
+4. Verify session-file behavior remains independent from skill sync.
+
+Success criteria:
+
+- Pi agents expose actual installed skill state
+- Paperclip can sync desired skills into Pi’s persistent home
+
+### 5.6 OpenCode Local
+
+Target mode:
+
+- `persistent`
+
+Special case:
+
+- OpenCode currently injects Paperclip skills into `~/.claude/skills`
+
+This is product-risky because:
+
+- it shares state with Claude
+- Paperclip may accidentally imply the skills belong only to OpenCode when the home is shared
+
+Plan:
+
+Phase 1:
+
+- implement `listSkills` and `syncSkills`
+- treat it as `persistent`
+- explicitly label the home as shared in UI copy
+- only remove stale managed Paperclip skills that are clearly marked as Paperclip-managed
+
+Phase 2:
+
+- investigate whether OpenCode supports its own isolated skills home
+- if yes, migrate to an adapter-specific home and remove the shared-home caveat
+
+Success criteria:
+
+- OpenCode agents show real state
+- shared-home risk is visible and bounded
+
+### 5.7 OpenClaw Gateway
+
+Target mode:
+
+- `unsupported` until gateway protocol support exists
+
+Required external work:
+
+- gateway API to list installed/available skills
+- gateway API to install/remove or otherwise reconcile skills
+- gateway metadata for whether state is persistent or ephemeral
+
+Until then:
+
+- Paperclip stores desired skills only
+- UI shows unsupported actual state
+- no fake sync implementation
+
+Future target:
+
+- likely a fourth truth model eventually, such as remote-managed persistent state
+- for now, keep the current API and treat gateway as unsupported
+
+## 6. API Plan
+
+## 6.1 Keep the current minimal adapter API
+
+Near-term adapter contract remains:
+
+- `listSkills(ctx)`
+- `syncSkills(ctx, desiredSkills)`
+
+This is enough for all local adapters.
+
+## 6.2 Optional extension points
+
+Add only if needed after the first broad rollout:
+
+- `skillHomeLabel`
+- `sharedHome: boolean`
+- `supportsExternalDiscovery: boolean`
+- `supportsDestructiveSync: boolean`
+
+These should be optional metadata additions to the snapshot, not required new adapter methods.
+
+## 7. UI Plan
+
+The company-level skill library can stay adapter-neutral.
+
+The agent-level Skills tab must become adapter-aware by copy and status:
+
+- `persistent`: installed / missing / stale / external
+- `ephemeral`: mounted on next run / external / desired only
+- `unsupported`: desired only, adapter cannot report actual state
+
+Additional UI requirement for shared-home adapters:
+
+- show a small warning that the adapter uses a shared user skills home
+- avoid destructive wording unless Paperclip can prove a skill is Paperclip-managed
+
+## 8. Rollout Phases
+
+### Phase 1: Finish the local filesystem family
+
+Ship:
+
+- `cursor_local`
+- `gemini_local`
+- `pi_local`
+
+Rationale:
+
+- these are the closest to Codex in architecture
+- they already inject into stable local skill homes
+
+### Phase 2: OpenCode shared-home support
+
+Ship:
+
+- `opencode_local`
+
+Rationale:
+
+- technically feasible now
+- needs slightly more careful product language because of the shared Claude skills home
+
+### Phase 3: Gateway support decision
+
+Decide:
+
+- keep `openclaw_gateway` unsupported for V1
+- or extend the gateway protocol for remote skill management
+
+My recommendation:
+
+- do not block V1 on gateway support
+- keep it explicitly unsupported until the remote protocol exists
+
+## 9. Definition Of Done
+
+Adapter-wide skill support is ready when all are true:
+
+1. Every adapter has an explicit truth model:
+   - `persistent`
+   - `ephemeral`
+   - `unsupported`
+2. The UI copy matches that truth model.
+3. All local persistent adapters implement:
+   - `listSkills`
+   - `syncSkills`
+4. Tests cover:
+   - desired-state storage
+   - actual-state discovery
+   - managed vs external distinctions
+   - stale managed-skill cleanup where supported
+5. `openclaw_gateway` is either:
+   - explicitly unsupported with clean UX
+   - or backed by a real remote skill API
+
+## 10. Recommendation
+
+The recommended immediate order is:
+
+1. `cursor_local`
+2. `gemini_local`
+3. `pi_local`
+4. `opencode_local`
+5. defer `openclaw_gateway`
+
+That gets Paperclip from “skills work for Codex and Claude” to “skills work for the whole local-adapter family,” which is the meaningful V1 milestone.

doc/plans/2026-03-14-billing-ledger-and-reporting.md

@@ -0,0 +1,468 @@
+# Billing Ledger and Reporting
+
+## Context
+
+Paperclip currently stores model spend in `cost_events` and operational run state in `heartbeat_runs`.
+That split is fine, but the current reporting code tries to infer billing semantics by mixing both tables:
+
+- `cost_events` knows provider, model, tokens, and dollars
+- `heartbeat_runs.usage_json` knows some per-run billing metadata
+- `heartbeat_runs.usage_json` does **not** currently carry enough normalized billing dimensions to support honest provider-level reporting
+
+This becomes incorrect as soon as a company uses more than one provider, more than one billing channel, or more than one billing mode.
+
+Examples:
+
+- direct OpenAI API usage
+- Claude subscription usage with zero marginal dollars
+- subscription overage with dollars and tokens
+- OpenRouter billing where the biller is OpenRouter but the upstream provider is Anthropic or OpenAI
+
+The system needs to support:
+
+- dollar reporting
+- token reporting
+- subscription-included usage
+- subscription overage
+- direct metered API usage
+- future aggregator billing such as OpenRouter
+
+## Product Decision
+
+`cost_events` becomes the canonical billing and usage ledger for reporting.
+
+`heartbeat_runs` remains an operational execution log. It may keep mirrored billing metadata for debugging and transcripts, but reporting must not reconstruct billing semantics from `heartbeat_runs.usage_json`.
+
+## Decision: One Ledger Or Two
+
+We do **not** need two tables to solve the current PR's problem.
+For request-level inference reporting, `cost_events` is enough if it carries the right dimensions:
+
+- upstream provider
+- biller
+- billing type
+- model
+- token fields
+- billed amount
+
+That is why the first implementation pass extends `cost_events` instead of introducing a second table immediately.
+
+However, if Paperclip needs to account for the full billing surface of aggregators and managed AI platforms, then `cost_events` alone is not enough.
+Some charges are not cleanly representable as a single model inference event:
+
+- account top-ups and credit purchases
+- platform fees charged at purchase time
+- BYOK platform fees that are account-level or threshold-based
+- prepaid credit expirations, refunds, and adjustments
+- provisioned throughput commitments
+- fine-tuning, training, model import, and storage charges
+- gateway logging or other platform overhead that is not attributable to one prompt/response pair
+
+So the decision is:
+
+- near term: keep `cost_events` as the inference and usage ledger
+- next phase: add `finance_events` for non-inference financial events
+
+This is a deliberate split between:
+
+- usage and inference accounting
+- account-level and platform-level financial accounting
+
+That separation keeps request reporting honest without forcing us to fake invoice semantics onto rows that were never request-scoped.
+
+## External Motivation And Sources
+
+The need for this model is not theoretical.
+It follows directly from the billing systems of providers and aggregators Paperclip needs to support.
+
+### OpenRouter
+
+Source URLs:
+
+- https://openrouter.ai/docs/faq#credit-and-billing-systems
+- https://openrouter.ai/pricing
+
+Relevant billing behavior as of March 14, 2026:
+
+- OpenRouter passes through underlying inference pricing and deducts request cost from purchased credits.
+- OpenRouter charges a 5.5% fee with a $0.80 minimum when purchasing credits.
+- Crypto payments are charged a 5% fee.
+- BYOK has its own fee model after a free request threshold.
+- OpenRouter billing is aggregated at the OpenRouter account level even when the upstream provider is Anthropic, OpenAI, Google, or another provider.
+
+Implication for Paperclip:
+
+- request usage belongs in `cost_events`
+- credit purchases, purchase fees, BYOK fees, refunds, and expirations belong in `finance_events`
+- `biller=openrouter` must remain distinct from `provider=anthropic|openai|google|...`
+
+### Cloudflare AI Gateway Unified Billing
+
+Source URL:
+
+- https://developers.cloudflare.com/ai-gateway/features/unified-billing/
+
+Relevant billing behavior as of March 14, 2026:
+
+- Unified Billing lets users call multiple upstream providers while receiving a single Cloudflare bill.
+- Usage is paid from Cloudflare-loaded credits.
+- Cloudflare supports manual top-ups and auto top-up thresholds.
+- Spend limits can stop request processing on daily, weekly, or monthly boundaries.
+- Unified Billing traffic can use Cloudflare-managed credentials rather than the user's direct provider key.
+
+Implication for Paperclip:
+
+- request usage needs `biller=cloudflare`
+- upstream provider still needs to be preserved separately
+- Cloudflare credit loads and related account-level events are not inference rows and should not be forced into `cost_events`
+- quota and limits reporting must support biller-level controls, not just upstream provider limits
+
+### Amazon Bedrock
+
+Source URL:
+
+- https://aws.amazon.com/bedrock/pricing/
+
+Relevant billing behavior as of March 14, 2026:
+
+- Bedrock supports on-demand and batch pricing.
+- Bedrock pricing varies by region.
+- some pricing tiers add premiums or discounts relative to standard pricing
+- provisioned throughput is commitment-based rather than request-based
+- custom model import uses Custom Model Units billed per minute, with monthly storage charges
+- imported model copies are billed in 5-minute windows once active
+- customization and fine-tuning introduce training and hosted-model charges beyond normal inference
+
+Implication for Paperclip:
+
+- normal tokenized inference fits in `cost_events`
+- provisioned throughput, custom model unit charges, training, and storage charges require `finance_events`
+- region and pricing tier need to be first-class dimensions in the financial model
+
+## Ledger Boundary
+
+To keep the system coherent, the table boundary should be explicit.
+
+### `cost_events`
+
+Use `cost_events` for request-scoped usage and inference charges:
+
+- one row per billable or usage-bearing run event
+- provider/model/biller/billingType/tokens/cost
+- optionally tied to `heartbeat_run_id`
+- supports direct APIs, subscriptions, overage, OpenRouter-routed inference, Cloudflare-routed inference, and Bedrock on-demand inference
+
+### `finance_events`
+
+Use `finance_events` for account-scoped or platform-scoped financial events:
+
+- credit purchase
+- top-up
+- refund
+- fee
+- expiry
+- provisioned capacity
+- training
+- model import
+- storage
+- invoice adjustment
+
+These rows may or may not have a related model, provider, or run id.
+Trying to force them into `cost_events` would either create fake request rows or create null-heavy rows that mean something fundamentally different from inference usage.
+
+## Canonical Billing Dimensions
+
+Every persisted billing event should model four separate axes:
+
+1. Usage provider
+   The upstream provider whose model performed the work.
+   Examples: `openai`, `anthropic`, `google`.
+
+2. Biller
+   The system that charged for the usage.
+   Examples: `openai`, `anthropic`, `openrouter`, `cursor`, `chatgpt`.
+
+3. Billing type
+   The pricing mode applied to the event.
+   Initial canonical values:
+   - `metered_api`
+   - `subscription_included`
+   - `subscription_overage`
+   - `credits`
+   - `fixed`
+   - `unknown`
+
+4. Measures
+   Usage and billing must both be storable:
+   - `input_tokens`
+   - `output_tokens`
+   - `cached_input_tokens`
+   - `cost_cents`
+
+These dimensions are independent.
+For example, an event may be:
+
+- provider: `anthropic`
+- biller: `openrouter`
+- billing type: `metered_api`
+- tokens: non-zero
+- cost cents: non-zero
+
+Or:
+
+- provider: `anthropic`
+- biller: `anthropic`
+- billing type: `subscription_included`
+- tokens: non-zero
+- cost cents: `0`
+
+## Schema Changes
+
+Extend `cost_events` with:
+
+- `heartbeat_run_id uuid null references heartbeat_runs.id`
+- `biller text not null default 'unknown'`
+- `billing_type text not null default 'unknown'`
+- `cached_input_tokens int not null default 0`
+
+Keep `provider` as the upstream usage provider.
+Do not overload `provider` to mean biller.
+
+Add a future `finance_events` table for account-level financial events with fields along these lines:
+
+- `company_id`
+- `occurred_at`
+- `event_kind`
+- `direction`
+- `biller`
+- `provider nullable`
+- `execution_adapter_type nullable`
+- `pricing_tier nullable`
+- `region nullable`
+- `model nullable`
+- `quantity nullable`
+- `unit nullable`
+- `amount_cents`
+- `currency`
+- `estimated`
+- `related_cost_event_id nullable`
+- `related_heartbeat_run_id nullable`
+- `external_invoice_id nullable`
+- `metadata_json nullable`
+
+Add indexes:
+
+- `(company_id, biller, occurred_at)`
+- `(company_id, provider, occurred_at)`
+- `(company_id, heartbeat_run_id)` if distinct-run reporting remains common
+
+## Shared Contract Changes
+
+### Shared types
+
+Add a shared billing type union and enrich cost types with:
+
+- `heartbeatRunId`
+- `biller`
+- `billingType`
+- `cachedInputTokens`
+
+Update reporting response types so the provider breakdown reflects the ledger directly rather than inferred run metadata.
+
+### Validators
+
+Extend `createCostEventSchema` to accept:
+
+- `heartbeatRunId`
+- `biller`
+- `billingType`
+- `cachedInputTokens`
+
+Defaults:
+
+- `biller` defaults to `provider`
+- `billingType` defaults to `unknown`
+- `cachedInputTokens` defaults to `0`
+
+## Adapter Contract Changes
+
+Extend adapter execution results so they can report:
+
+- `biller`
+- richer billing type values
+
+Backwards compatibility:
+
+- existing adapter values `api` and `subscription` are treated as legacy aliases
+- map `api -> metered_api`
+- map `subscription -> subscription_included`
+
+Future adapters may emit the canonical values directly.
+
+OpenRouter support will use:
+
+- `provider` = upstream provider when known
+- `biller` = `openrouter`
+- `billingType` = `metered_api` unless OpenRouter later exposes another billing mode
+
+Cloudflare Unified Billing support will use:
+
+- `provider` = upstream provider when known
+- `biller` = `cloudflare`
+- `billingType` = `credits` or `metered_api` depending on the normalized request billing contract
+
+Bedrock support will use:
+
+- `provider` = upstream provider or `aws_bedrock` depending on adapter shape
+- `biller` = `aws_bedrock`
+- `billingType` = request-scoped mode for inference rows
+- `finance_events` for provisioned, training, import, and storage charges
+
+## Write Path Changes
+
+### Heartbeat-created events
+
+When a heartbeat run produces usage or spend:
+
+1. normalize adapter billing metadata
+2. write a ledger row to `cost_events`
+3. attach `heartbeat_run_id`
+4. set `provider`, `biller`, `billing_type`, token fields, and `cost_cents`
+
+The write path should no longer depend on later inference from `heartbeat_runs`.
+
+### Manual API-created events
+
+Manual cost event creation remains supported.
+These events may have `heartbeatRunId = null`.
+
+Rules:
+
+- `provider` remains required
+- `biller` defaults to `provider`
+- `billingType` defaults to `unknown`
+
+## Reporting Changes
+
+### Server
+
+Refactor reporting queries to use `cost_events` only.
+
+#### `summary`
+
+- sum `cost_cents`
+
+#### `by-agent`
+
+- sum costs and token fields from `cost_events`
+- use `count(distinct heartbeat_run_id)` filtered by billing type for run counts
+- use token sums filtered by billing type for subscription usage
+
+#### `by-provider`
+
+- group by `provider`, `model`
+- sum costs and token fields directly from the ledger
+- derive billing-type slices from `cost_events.billing_type`
+- never pro-rate from unrelated `heartbeat_runs`
+
+#### future `by-biller`
+
+- group by `biller`
+- this is the right view for invoice and subscription accountability
+
+#### `window-spend`
+
+- continue to use `cost_events`
+
+#### project attribution
+
+Keep current project attribution logic for now, but prefer `cost_events.heartbeat_run_id` as the join anchor whenever possible.
+
+## UI Changes
+
+### Principles
+
+- Spend, usage, and quota are related but distinct
+- a missing quota fetch is not the same as “no quota”
+- provider and biller are different dimensions
+
+### Immediate UI changes
+
+1. Keep the current costs page structure.
+2. Make the provider cards accurate by reading only ledger-backed values.
+3. Show provider quota fetch errors explicitly instead of dropping them.
+
+### Follow-up UI direction
+
+The long-term board UI should expose:
+
+- Spend
+  Dollars by biller, provider, model, agent, project
+- Usage
+  Tokens by provider, model, agent, project
+- Quotas
+  Live provider or biller limits, credits, and reset windows
+- Financial events
+  Credit purchases, top-ups, fees, refunds, commitments, storage, and other non-inference charges
+
+## Migration Plan
+
+Migration behavior:
+
+- add new non-destructive columns with defaults
+- backfill existing rows:
+  - `biller = provider`
+  - `billing_type = 'unknown'`
+  - `cached_input_tokens = 0`
+  - `heartbeat_run_id = null`
+
+Do **not** attempt to backfill historical provider-level subscription attribution from `heartbeat_runs`.
+That data was never stored with the required dimensions.
+
+## Testing Plan
+
+Add or update tests for:
+
+1. heartbeat-created ledger rows persist `heartbeatRunId`, `biller`, `billingType`, and cached tokens
+2. legacy adapter billing values map correctly
+3. provider reporting uses ledger data only
+4. mixed-provider companies do not cross-attribute subscription usage
+5. zero-dollar subscription usage still appears in token reporting
+6. quota fetch failures render explicit UI state
+7. manual cost events still validate and write correctly
+8. biller reporting keeps upstream provider breakdowns separate
+9. OpenRouter-style rows can show `biller=openrouter` with non-OpenRouter upstream providers
+10. Cloudflare-style rows can show `biller=cloudflare` with preserved upstream provider identity
+11. future `finance_events` aggregation handles non-request charges without requiring a model or run id
+
+## Delivery Plan
+
+### Step 1
+
+- land the ledger contract and query rewrite
+- make the current costs page correct
+
+### Step 2
+
+- add biller-oriented reporting endpoints and UI
+
+### Step 3
+
+- wire OpenRouter and any future aggregator adapters to the same contract
+
+### Step 4
+
+- add `executionAdapterType` to persisted cost reporting if adapter-level grouping becomes a product requirement
+
+### Step 5
+
+- introduce `finance_events`
+- add non-inference accounting endpoints
+- add UI for platform/account charges alongside inference spend and usage
+
+## Non-Goals For This Change
+
+- multi-currency support
+- invoice reconciliation
+- provider-specific cost estimation beyond persisted billed cost
+- replacing `heartbeat_runs` as the operational run record

doc/plans/2026-03-14-budget-policies-and-enforcement.md

@@ -0,0 +1,611 @@
+# Budget Policies and Enforcement
+
+## Context
+
+Paperclip already treats budgets as a core control-plane responsibility:
+
+- `doc/SPEC.md` gives the Board authority to set budgets, pause agents, pause work, and override any budget.
+- `doc/SPEC-implementation.md` says V1 must support monthly UTC budget windows, soft alerts, and hard auto-pause.
+- the current code only partially implements that intent.
+
+Today the system has narrow money-budget behavior:
+
+- companies track `budgetMonthlyCents` and `spentMonthlyCents`
+- agents track `budgetMonthlyCents` and `spentMonthlyCents`
+- `cost_events` ingestion increments those counters
+- when an agent exceeds its monthly budget, the agent is paused
+
+That leaves major product gaps:
+
+- no project budget model
+- no approval generated when budget is hit
+- no generic budget policy system
+- no project pause semantics tied to budget
+- no durable incident tracking to prevent duplicate alerts
+- no separation between enforceable spend budgets and advisory usage quotas
+
+This plan defines the concrete budgeting model Paperclip should implement next.
+
+## Product Goals
+
+Paperclip should let operators:
+
+1. Set budgets on agents and projects.
+2. Understand whether a budget is based on money or usage.
+3. Be warned before a budget is exhausted.
+4. Automatically pause work when a hard budget is hit.
+5. Approve, raise, or resume from a budget stop using obvious UI.
+6. See budget state on the dashboard, `/costs`, and scope detail pages.
+
+The system should make one thing very clear:
+
+- budgets are policy controls
+- quotas are usage visibility
+
+They are related, but they are not the same concept.
+
+## Product Decisions
+
+### V1 Budget Defaults
+
+For the next implementation pass, Paperclip should enforce these defaults:
+
+- agent budgets are recurring monthly budgets
+- project budgets are lifetime total budgets
+- hard-stop enforcement uses billed dollars, not tokens
+- monthly windows use UTC calendar months
+- project total budgets do not reset automatically
+
+This gives a clean mental model:
+
+- agents are ongoing workers, so monthly recurring budget is natural
+- projects are bounded workstreams, so lifetime cap is natural
+
+### Metric To Enforce First
+
+The first enforceable metric should be `billed_cents`.
+
+Reasoning:
+
+- it works across providers, billers, and models
+- it maps directly to real financial risk
+- it handles overage and metered usage consistently
+- it avoids cross-provider token normalization problems
+- it applies cleanly even when future finance events are not token-based
+
+Token budgets should not be the first hard-stop policy.
+They should come later as advisory usage controls once the money-based system is solid.
+
+### Subscription Usage Decision
+
+Paperclip should separate subscription-included usage from billed spend:
+
+- `subscription_included`
+  - visible in reporting
+  - visible in usage summaries
+  - does not count against money budget
+- `subscription_overage`
+  - visible in reporting
+  - counts against money budget
+- `metered_api`
+  - visible in reporting
+  - counts against money budget
+
+This keeps the budget system honest:
+
+- users should not see "spend" rise for usage that did not incur marginal billed cost
+- users should still see the token usage and provider quota state
+
+### Soft Alert Versus Hard Stop
+
+Paperclip should have two threshold classes:
+
+- soft alert
+  - creates visible notification state
+  - does not create an approval
+  - does not pause work
+- hard stop
+  - pauses the affected scope automatically
+  - creates an approval requiring human resolution
+  - prevents additional heartbeats or task pickup in that scope
+
+Default thresholds:
+
+- soft alert at `80%`
+- hard stop at `100%`
+
+These should be configurable per policy later, but they are good defaults now.
+
+## Scope Model
+
+### Supported Scope Types
+
+Budget policies should support:
+
+- `company`
+- `agent`
+- `project`
+
+This plan focuses on finishing `agent` and `project` first while preserving the existing company budget behavior.
+
+### Recommended V1.5 Policy Presets
+
+- Company
+  - metric: `billed_cents`
+  - window: `calendar_month_utc`
+- Agent
+  - metric: `billed_cents`
+  - window: `calendar_month_utc`
+- Project
+  - metric: `billed_cents`
+  - window: `lifetime`
+
+Future extensions can add:
+
+- token advisory policies
+- daily or weekly spend windows
+- provider- or biller-scoped budgets
+- inherited delegated budgets down the org tree
+
+## Current Implementation Baseline
+
+The current codebase is not starting from zero, but the existing shape is too ad hoc to extend safely.
+
+### What Exists Today
+
+- company and agent monthly cents counters
+- cost ingestion that updates those counters
+- agent hard-stop pause on monthly budget overrun
+
+### What Is Missing
+
+- project budgets
+- generic budget policy persistence
+- generic threshold crossing detection
+- incident deduplication per scope/window
+- approval creation on hard-stop
+- project execution blocking
+- budget timeline and incident UI
+- distinction between advisory quota and enforceable budget
+
+## Proposed Data Model
+
+### 1. `budget_policies`
+
+Create a new table for canonical budget definitions.
+
+Suggested fields:
+
+- `id`
+- `company_id`
+- `scope_type`
+- `scope_id`
+- `metric`
+- `window_kind`
+- `amount`
+- `warn_percent`
+- `hard_stop_enabled`
+- `notify_enabled`
+- `is_active`
+- `created_by_user_id`
+- `updated_by_user_id`
+- `created_at`
+- `updated_at`
+
+Notes:
+
+- `scope_type` is one of `company | agent | project`
+- `scope_id` is nullable only for company-level policy if company is implied; otherwise keep it explicit
+- `metric` should start with `billed_cents`
+- `window_kind` starts with `calendar_month_utc | lifetime`
+- `amount` is stored in the natural unit of the metric
+
+### 2. `budget_incidents`
+
+Create a durable record of threshold crossings.
+
+Suggested fields:
+
+- `id`
+- `company_id`
+- `policy_id`
+- `scope_type`
+- `scope_id`
+- `metric`
+- `window_kind`
+- `window_start`
+- `window_end`
+- `threshold_type`
+- `amount_limit`
+- `amount_observed`
+- `status`
+- `approval_id` nullable
+- `activity_id` nullable
+- `resolved_at` nullable
+- `created_at`
+- `updated_at`
+
+Notes:
+
+- `threshold_type`: `soft | hard`
+- `status`: `open | acknowledged | resolved | dismissed`
+- one open incident per policy per threshold per window prevents duplicate approvals and alert spam
+
+### 3. Project Pause State
+
+Projects need explicit pause semantics.
+
+Recommended approach:
+
+- extend project status or add a pause field so a project can be blocked by budget
+- preserve whether the project is paused due to budget versus manually paused
+
+Preferred shape:
+
+- keep project workflow status as-is
+- add execution-state fields:
+  - `execution_status`: `active | paused | archived`
+  - `pause_reason`: `manual | budget | system | null`
+
+If that is too large for the immediate pass, a smaller version is:
+
+- add `paused_at`
+- add `pause_reason`
+
+The key requirement is behavioral, not cosmetic:
+Paperclip must know that a project is budget-paused and enforce it.
+
+### 4. Compatibility With Existing Budget Columns
+
+Existing company and agent monthly budget columns should remain temporarily for compatibility.
+
+Migration plan:
+
+1. keep reading existing columns during transition
+2. create equivalent `budget_policies` rows
+3. switch enforcement and UI to policies
+4. later remove or deprecate legacy columns
+
+## Budget Engine
+
+Budget enforcement should move into a dedicated service.
+
+Current logic is buried inside cost ingestion.
+That is too narrow because budget checks must apply at more than one execution boundary.
+
+### Responsibilities
+
+New service: `budgetService`
+
+Responsibilities:
+
+- resolve applicable policies for a cost event
+- compute current window totals
+- detect threshold crossings
+- create incidents, activities, and approvals
+- pause affected scopes on hard-stop
+- provide preflight enforcement checks for execution entry points
+
+### Canonical Evaluation Flow
+
+When a new `cost_event` is written:
+
+1. persist the `cost_event`
+2. identify affected scopes
+   - company
+   - agent
+   - project
+3. fetch active policies for those scopes
+4. compute current observed amount for each policy window
+5. compare to thresholds
+6. create soft incident if soft threshold crossed for first time in window
+7. create hard incident if hard threshold crossed for first time in window
+8. if hard incident:
+   - pause the scope
+   - create approval
+   - create activity event
+   - emit notification state
+
+### Preflight Enforcement Checks
+
+Budget enforcement cannot rely only on post-hoc cost ingestion.
+
+Paperclip must also block execution before new work starts.
+
+Add budget checks to:
+
+- scheduler heartbeat dispatch
+- manual invoke endpoints
+- assignment-driven wakeups
+- queued run promotion
+- issue checkout or pickup paths where applicable
+
+If a scope is budget-paused:
+
+- do not start a new heartbeat
+- do not let the agent pick up additional work
+- present a clear reason in API and UI
+
+### Active Run Behavior
+
+When a hard-stop is triggered while a run is already active:
+
+- mark scope paused immediately for future work
+- request graceful cancellation of the current run
+- allow normal cancellation timeout behavior
+- write activity explaining that pause came from budget enforcement
+
+This mirrors the general pause semantics already expected by the product.
+
+## Approval Model
+
+Budget hard-stops should create a first-class approval.
+
+### New Approval Type
+
+Add approval type:
+
+- `budget_override_required`
+
+Payload should include:
+
+- `scopeType`
+- `scopeId`
+- `scopeName`
+- `metric`
+- `windowKind`
+- `thresholdType`
+- `budgetAmount`
+- `observedAmount`
+- `windowStart`
+- `windowEnd`
+- `topDrivers`
+- `paused`
+
+### Resolution Actions
+
+The approval UI should support:
+
+- raise budget and resume
+- resume once without changing policy
+- keep paused
+
+Optional later action:
+
+- disable budget policy
+
+### Soft Alerts Do Not Need Approval
+
+Soft alerts should create:
+
+- activity event
+- dashboard alert
+- inbox notification or similar board-visible signal
+
+They should not create an approval by default.
+
+## Notification And Activity Model
+
+Budget events need obvious operator visibility.
+
+Required outputs:
+
+- activity log entry on threshold crossings
+- dashboard surface for active budget incidents
+- detail page banner on paused agent or project
+- `/costs` summary of active incidents and policy health
+
+Later channels:
+
+- email
+- webhook
+- Slack or other integrations
+
+## API Plan
+
+### Policy Management
+
+Add routes for:
+
+- list budget policies for company
+- create budget policy
+- update budget policy
+- archive or disable budget policy
+
+### Incident Surfaces
+
+Add routes for:
+
+- list active budget incidents
+- list incident history
+- get incident detail for a scope
+
+### Approval Resolution
+
+Budget approvals should use the existing approval system once the new approval type is added.
+
+Expected flows:
+
+- create approval on hard-stop
+- resolve approval by changing policy and resuming
+- resolve approval by resuming once
+
+### Execution Errors
+
+When work is blocked by budget, the API should return explicit errors.
+
+Examples:
+
+- agent invocation blocked because agent budget is paused
+- issue execution blocked because project budget is paused
+
+Do not silently no-op.
+
+## UI Plan
+
+Budgeting should be visible in the places where operators make decisions.
+
+### `/costs`
+
+Add a budget section that includes:
+
+- active budget incidents
+- policy list with scope, window, metric, and threshold state
+- progress bars for current period or total
+- clear distinction between:
+  - spend budget
+  - subscription quota
+- quick actions:
+  - raise budget
+  - open approval
+  - resume scope if permitted
+
+The page should make this visual distinction obvious:
+
+- Budget
+  - enforceable spend policy
+- Quota
+  - provider or subscription usage window
+
+### Agent Detail
+
+Add an agent budget card:
+
+- monthly budget amount
+- current month spend
+- remaining spend
+- status
+- warning or paused banner
+- link to approval if blocked
+
+### Project Detail
+
+Add a project budget card:
+
+- total budget amount
+- total spend to date
+- remaining spend
+- pause status
+- approval link
+
+Project detail should also show if issue execution is blocked because the project is budget-paused.
+
+### Dashboard
+
+Add a high-signal budget section:
+
+- active budget breaches
+- upcoming soft alerts
+- counts of paused agents and paused projects due to budget
+
+The operator should not have to visit `/costs` to learn that work has stopped.
+
+## Budget Math
+
+### What Counts Toward Budget
+
+For V1.5 enforcement, include:
+
+- `metered_api` cost events
+- `subscription_overage` cost events
+- any future request-scoped cost event with non-zero billed cents
+
+Do not include:
+
+- `subscription_included` cost events with zero billed cents
+- advisory quota rows
+- account-level finance events unless and until company-level financial budgets are added explicitly
+
+### Why Not Tokens First
+
+Token budgets should not be the first hard-stop because:
+
+- providers count tokens differently
+- cached tokens complicate simple totals
+- some future charges are not token-based
+- subscription tokens do not necessarily imply spend
+- money remains the cleanest cross-provider enforcement metric
+
+### Future Budget Metrics
+
+Future policy metrics can include:
+
+- `total_tokens`
+- `input_tokens`
+- `output_tokens`
+- `requests`
+- `finance_amount_cents`
+
+But they should enter only after the money-budget path is stable.
+
+## Migration Plan
+
+### Phase 1: Foundation
+
+- add `budget_policies`
+- add `budget_incidents`
+- add new approval type
+- add project pause metadata
+
+### Phase 2: Compatibility
+
+- backfill policies from existing company and agent monthly budget columns
+- keep legacy columns readable during migration
+
+### Phase 3: Enforcement
+
+- move budget logic into dedicated service
+- add hard-stop incident creation
+- add activity and approval creation
+- add execution guards on heartbeat and invoke paths
+
+### Phase 4: UI
+
+- `/costs` budget section
+- agent detail budget card
+- project detail budget card
+- dashboard incident summary
+
+### Phase 5: Cleanup
+
+- move all reads/writes to `budget_policies`
+- reduce legacy column reliance
+- decide whether to remove old budget columns
+
+## Tests
+
+Required coverage:
+
+- agent monthly budget soft alert at 80%
+- agent monthly budget hard-stop at 100%
+- project lifetime budget soft alert
+- project lifetime budget hard-stop
+- `subscription_included` usage does not consume money budget
+- `subscription_overage` does consume money budget
+- hard-stop creates one incident per threshold per window
+- hard-stop creates approval and pauses correct scope
+- paused project blocks new issue execution
+- paused agent blocks new heartbeat dispatch
+- policy update and resume clears or resolves active incident correctly
+- dashboard and `/costs` surface active incidents
+
+## Open Questions
+
+These should be explicitly deferred unless they block implementation:
+
+- Should project budgets also support monthly mode, or is lifetime enough for the first release?
+- Should company-level budgets eventually include `finance_events` such as OpenRouter top-up fees and Bedrock provisioned charges?
+- Should delegated budget editing be limited by org hierarchy in V1, or remain board-only in the UI even if the data model can support delegation later?
+- Do we need "resume once" immediately, or can first approval resolution be "raise budget and resume" plus "keep paused"?
+
+## Recommendation
+
+Implement the first coherent budgeting system with these rules:
+
+- Agent budget = monthly billed dollars
+- Project budget = lifetime billed dollars
+- Hard-stop = auto-pause + approval
+- Soft alert = visible warning, no approval
+- Subscription usage = visible quota and token reporting, not money-budget enforcement
+
+This solves the real operator problem without mixing together spend control, provider quota windows, and token accounting.