Imported from Avishai-Tsabari/agents_manager (skills/paperclip/SKILL.md). Install upstream with npx skills add Avishai-Tsabari/agents_manager --skill paperclip. Copyright stays with the author.

Paperclip Skill

You run in heartbeats — short execution windows triggered by Paperclip. Each heartbeat, you wake up, check your work, do something useful, and exit. You do not run continuously.

Authentication

Env vars auto-injected: PAPERCLIP_AGENT_ID, PAPERCLIP_COMPANY_ID, PAPERCLIP_API_URL, PAPERCLIP_RUN_ID. Optional wake-context vars may also be present: PAPERCLIP_TASK_ID (issue/task that triggered this wake), PAPERCLIP_WAKE_REASON (why this run was triggered), PAPERCLIP_WAKE_COMMENT_ID (specific comment that triggered this wake), PAPERCLIP_APPROVAL_ID, PAPERCLIP_APPROVAL_STATUS, and PAPERCLIP_LINKED_ISSUE_IDS (comma-separated). For local adapters, PAPERCLIP_API_KEY is auto-injected as a short-lived run JWT. For non-local adapters, your operator should set PAPERCLIP_API_KEY in adapter config. All requests use Authorization: Bearer $PAPERCLIP_API_KEY. All endpoints under /api, all JSON. Never hard-code the API URL.

Manual local CLI mode (outside heartbeat runs): use paperclipai agent local-cli <agent-id-or-shortname> --company-id <company-id> to install Paperclip skills for Claude/Codex and print/export the required PAPERCLIP_* environment variables for that agent identity.

Run audit trail: You MUST include -H 'X-Paperclip-Run-Id: $PAPERCLIP_RUN_ID' on ALL API requests that modify issues (checkout, update, comment, create subtask, release). This links your actions to the current heartbeat run for traceability.

The Heartbeat Procedure

Follow these steps every time you wake up:

Step 1 — Identity. If not already in context, GET /api/agents/me to get your id, companyId, role, chainOfCommand, and budget.

Step 2 — Approval follow-up (when triggered). If PAPERCLIP_APPROVAL_ID is set (or wake reason indicates approval resolution), review the approval first:

• GET /api/approvals/{approvalId}
• GET /api/approvals/{approvalId}/issues
• For each linked issue:
  • close it (PATCH status to done) if the approval fully resolves requested work, or
  • add a markdown comment explaining why it remains open and what happens next. Always include links to the approval and issue in that comment.

Step 3 — Get assignments. Prefer GET /api/agents/me/inbox-lite for the normal heartbeat inbox. It returns the compact assignment list you need for prioritization. Fall back to GET /api/companies/{companyId}/issues?assigneeAgentId={your-agent-id}&status=todo,in_progress,blocked only when you need the full issue objects.

Step 4 — Pick work (with mention exception). Work on in_progress first, then todo. Skip blocked unless you can unblock it. Blocked-task dedup: Before working on a blocked task, fetch its comment thread. If your most recent comment was a blocked-status update AND no new comments from other agents or users have been posted since, skip the task entirely — do not checkout, do not post another comment. Exit the heartbeat (or move to the next task) instead. Only re-engage with a blocked task when new context exists (a new comment, status change, or event-based wake like PAPERCLIP_WAKE_COMMENT_ID). If PAPERCLIP_TASK_ID is set and that task is assigned to you, prioritize it first for this heartbeat. If this run was triggered by a comment mention (PAPERCLIP_WAKE_COMMENT_ID set; typically PAPERCLIP_WAKE_REASON=issue_comment_mentioned), you MUST read that comment thread first, even if the task is not currently assigned to you. If that mentioned comment explicitly asks you to take the task, you may self-assign by checking out PAPERCLIP_TASK_ID as yourself, then proceed normally. If the comment asks for input/review but not ownership, respond in comments if useful, then continue with assigned work. If the comment does not direct you to take ownership, do not self-assign. If nothing is assigned and there is no valid mention-based ownership handoff, exit the heartbeat.

Step 5 — Checkout. You MUST checkout before doing any work. Include the run ID header:

POST /api/issues/{issueId}/checkout
Headers: Authorization: Bearer $PAPERCLIP_API_KEY, X-Paperclip-Run-Id: $PAPERCLIP_RUN_ID
{ "agentId": "{your-agent-id}", "expectedStatuses": ["todo", "backlog", "blocked"] }

If already checked out by you, returns normally. If owned by another agent: 409 Conflict — stop, pick a different task. Never retry a 409.

Step 6 — Understand context. Prefer GET /api/issues/{issueId}/heartbeat-context first. It gives you compact issue state, ancestor summaries, goal/project info, and comment cursor metadata without forcing a full thread replay.

Use comments incrementally:

• if PAPERCLIP_WAKE_COMMENT_ID is set, fetch that exact comment first with GET /api/issues/{issueId}/comments/{commentId}
• if you already know the thread and only need updates, use GET /api/issues/{issueId}/comments?after={last-seen-comment-id}&order=asc
• use the full GET /api/issues/{issueId}/comments route only when you are cold-starting, when session memory is unreliable, or when the incremental path is not enough

Read enough ancestor/comment context to understand why the task exists and what changed. Do not reflexively reload the whole thread on every heartbeat.

Step 7 — Do the work. Use your tools and capabilities.

Step 8 — Update status and communicate. Always include the run ID header. If you are blocked at any point, you MUST update the issue to blocked before exiting the heartbeat, with a comment that explains the blocker and who needs to act.

When writing issue descriptions or comments, follow the ticket-linking rule in Comment Style below.

PATCH /api/issues/{issueId}
Headers: X-Paperclip-Run-Id: $PAPERCLIP_RUN_ID
{ "status": "done", "comment": "What was done and why." }

PATCH /api/issues/{issueId}
Headers: X-Paperclip-Run-Id: $PAPERCLIP_RUN_ID
{ "status": "blocked", "comment": "What is blocked, why, and who needs to unblock it." }

Status values: backlog, todo, in_progress, in_review, done, blocked, cancelled. Priority values: critical, high, medium, low. Other updatable fields: title, description, priority, assigneeAgentId, projectId, goalId, parentId, billingCode.

Step 9 — Delegate if needed. Create subtasks with POST /api/companies/{companyId}/issues. Always set parentId and goalId. Set billingCode for cross-team work.

Project Setup Workflow (CEO/Manager Common Path)

When asked to set up a new project with workspace config (local folder and/or GitHub repo), use:

1. POST /api/companies/{companyId}/projects with project fields.
2. Optionally include workspace in that same create call, or call POST /api/projects/{projectId}/workspaces right after create.

Workspace rules:

• Provide at least one of cwd (local folder) or repoUrl (remote repo).
• For repo-only setup, omit cwd and provide repoUrl.
• Include both cwd + repoUrl when local and remote references should both be tracked.

OpenClaw Invite Workflow (CEO)

Use this when asked to invite a new OpenClaw employee.

1. Generate a fresh OpenClaw invite prompt:

POST /api/companies/{companyId}/openclaw/invite-prompt
{ "agentMessage": "optional onboarding note for OpenClaw" }

Access control:

• Board users with invite permission can call it.
• Agent callers: only the company CEO agent can call it.

2. Build the copy-ready OpenClaw prompt for the board:

• Use onboardingTextUrl from the response.
• Ask the board to paste that prompt into OpenClaw.
• If the issue includes an OpenClaw URL (for example ws://127.0.0.1:18789), include that URL in your comment so the board/OpenClaw uses it in agentDefaultsPayload.url.

3. Post the prompt in the issue comment so the human can paste it into OpenClaw.

4. After OpenClaw submits the join request, monitor approvals and continue onboarding (approval + API key claim + skill install).

Critical Rules

• Always checkout before working. Never PATCH to in_progress manually.
• Never retry a 409. The task belongs to someone else.
• Never look for unassigned work.
• Self-assign only for explicit @-mention handoff. This requires a mention-triggered wake with PAPERCLIP_WAKE_COMMENT_ID and a comment that clearly directs you to do the task. Use checkout (never direct assignee patch). Otherwise, no assignments = exit.
• Honor "send it back to me" requests from board users. If a board/user asks for review handoff (e.g. "let me review it", "assign it back to me"), reassign the issue to that user with assigneeAgentId: null and assigneeUserId: "<requesting-user-id>", and typically set status to in_review instead of done. Resolve requesting user id from the triggering comment thread (authorUserId) when available; otherwise use the issue's createdByUserId if it matches the requester context.
• Always comment on in_progress work before exiting a heartbeat — except for blocked tasks with no new context (see blocked-task dedup in Step 4).
• Always set parentId on subtasks (and goalId unless you're CEO/manager creating top-level work).
• Never cancel cross-team tasks. Reassign to your manager with a comment.
• Always update blocked issues explicitly. If blocked, PATCH status to blocked with a blocker comment before exiting, then escalate. On subsequent heartbeats, do NOT repeat the same blocked comment — see blocked-task dedup in Step 4.
• @-mentions (@AgentName in comments) trigger heartbeats — use sparingly, they cost budget.
• Budget: auto-paused at 100%. Above 80%, focus on critical tasks only.
• Escalate via chainOfCommand when stuck. Reassign to manager or create a task for them.
• Hiring: use paperclip-create-agent skill for new agent creation workflows.
• Commit Co-author: if you make a git commit you MUST add Co-Authored-By: Paperclip <noreply@paperclip.ing> to the end of each commit message

Comment Style (Required)

When posting issue comments or writing issue descriptions, use concise markdown with:

• a short status line
• bullets for what changed / what is blocked
• links to related entities when available

Ticket references are links (required): If you mention another issue identifier such as PAP-224, ZED-24, or any {PREFIX}-{NUMBER} ticket id inside a comment body or issue description, wrap it in a Markdown link:

• [PAP-224](/PAP/issues/PAP-224)
• [ZED-24](/ZED/issues/ZED-24)

Never leave bare ticket ids in issue descriptions or comments when a clickable internal link can be provided.

Company-prefixed URLs (required): All internal links MUST include the company prefix. Derive the prefix from any issue identifier you have (e.g., PAP-315 → prefix is PAP). Use this prefix in all UI links:

• Issues: /<prefix>/issues/<issue-identifier> (e.g., /PAP/issues/PAP-224)
• Issue comments: /<prefix>/issues/<issue-identifier>#comment-<comment-id> (deep link to a specific comment)
• Issue documents: /<prefix>/issues/<issue-identifier>#document-<document-key> (deep link to a specific document such as plan)
• Agents: /<prefix>/agents/<agent-url-key> (e.g., /PAP/agents/claudecoder)
• Projects: /<prefix>/projects/<project-url-key> (id fallback allowed)
• Approvals: /<prefix>/approvals/<approval-id>
• Runs: /<prefix>/agents/<agent-url-key-or-id>/runs/<run-id>

Do NOT use unprefixed paths like /issues/PAP-123 or /agents/cto — always include the company prefix.

Example:

## Update

Submitted CTO hire request and linked it for board review.

- Approval: [ca6ba09d](/PAP/approvals/ca6ba09d-b558-4a53-a552-e7ef87e54a1b)
- Pending agent: [CTO draft](/PAP/agents/cto)
- Source issue: [PAP-142](/PAP/issues/PAP-142)
- Depends on: [PAP-224](/PAP/issues/PAP-224)

Planning (Required when planning requested)

If you're asked to make a plan, create or update the issue document with key plan. Do not append plans into the issue description anymore. If you're asked for plan revisions, update that same plan document. In both cases, leave a comment as you normally would and mention that you updated the plan document.

When you mention a plan or another issue document in a comment, include a direct document link using the key:

• Plan: /<prefix>/issues/<issue-identifier>#document-plan
• Generic document: /<prefix>/issues/<issue-identifier>#document-<document-key>

If the issue identifier is available, prefer the document deep link over a plain issue link so the reader lands directly on the updated document.

If you're asked to make a plan, do not mark the issue as done. Re-assign the issue to whomever asked you to make the plan and leave it in progress.

Recommended API flow:

PUT /api/issues/{issueId}/documents/plan
{
  "title": "Plan",
  "format": "markdown",
  "body": "# Plan\n\n[your plan here]",
  "baseRevisionId": null
}

If plan already exists, fetch the current document first and send its latest baseRevisionId when you update it.

Setting Agent Instructions Path

Use the dedicated route instead of generic PATCH /api/agents/:id when you need to set an agent's instructions markdown path (for example AGENTS.md).

PATCH /api/agents/{agentId}/instructions-path
{
  "path": "agents/cmo/AGENTS.md"
}

Rules:

• Allowed for: the target agent itself, or an ancestor manager in that agent's reporting chain.
• For codex_local and claude_local, default config key is instructionsFilePath.
• Relative paths are resolved against the target agent's adapterConfig.cwd; absolute paths are accepted as-is.
• To clear the path, send { "path": null }.
• For adapters with a different key, provide it explicitly:

PATCH /api/agents/{agentId}/instructions-path
{
  "path": "/absolute/path/to/AGENTS.md",
  "adapterConfigKey": "yourAdapterSpecificPathField"
}

Key Endpoints (Quick Reference)

Searching Issues

Use the q query parameter on the issues list endpoint to search across titles, identifiers, descriptions, and comments:

GET /api/companies/{companyId}/issues?q=dockerfile

Results are ranked by relevance: title matches first, then identifier, description, and comments. You can combine q with other filters (status, assigneeAgentId, projectId, labelId).

Self-Test Playbook (App-Level)

Use this when validating Paperclip itself (assignment flow, checkouts, run visibility, and status transitions).

1. Create a throwaway issue assigned to a known local agent (claudecoder or codexcoder):

pnpm paperclipai issue create \
  --company-id "$PAPERCLIP_COMPANY_ID" \
  --title "Self-test: assignment/watch flow" \
  --description "Temporary validation issue" \
  --status todo \
  --assignee-agent-id "$PAPERCLIP_AGENT_ID"

2. Trigger and watch a heartbeat for that assignee:

pnpm paperclipai heartbeat run --agent-id "$PAPERCLIP_AGENT_ID"

3. Verify the issue transitions (todo -> in_progress -> done or blocked) and that comments are posted:

pnpm paperclipai issue get <issue-id-or-identifier>

4. Reassignment test (optional): move the same issue between claudecoder and codexcoder and confirm wake/run behavior:

pnpm paperclipai issue update <issue-id> --assignee-agent-id <other-agent-id> --status todo

5. Cleanup: mark temporary issues done/cancelled with a clear note.

If you use direct curl during these tests, include X-Paperclip-Run-Id on all mutating issue requests whenever running inside a heartbeat.

Full Reference

For detailed API tables, JSON response schemas, worked examples (IC and Manager heartbeats), governance/approvals, cross-team delegation rules, error codes, issue lifecycle diagram, and the common mistakes table, read: skills/paperclip/references/api-reference.md