Skill

Autopilot

Backend-managed autopilot poller that claims kanban cards from autopilot-enabled columns and dispatches agent runs

npx claudepluginhub pragnition/pragnition-public-plugins --plugin rapid

Tool Access

This skill cannot use any tools. It operates in read-only mode without the ability to modify files or execute commands.

Preview

The autopilot is a backend-managed background worker, not a CLI-invoked skill. It runs as a lifespan-managed poller inside the RAPID web backend (`app/agents/autopilot_worker.py`).

SKILL.md

Similar Skills

skill-lookup

159.9k

Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.

prompts.chat

prompt-lookup

159.9k

Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.

prompts.chat

next-compile

139.2k

Checks Next.js compilation errors using a running Turbopack dev server after code edits. Fixes actionable issues before reporting complete. Replaces `next build`.

1 file

vercel-next-js-2

Stats

Stars4

Forks0

Last CommitApr 17, 2026

Actions

View Source View Plugin View on GitHub View README

Autopilot

The autopilot is a backend-managed background worker, not a CLI-invoked skill. It runs as a lifespan-managed poller inside the RAPID web backend (app/agents/autopilot_worker.py).

How It Works

Polls autopilot-enabled columns for unclaimed kanban cards on a 60-second interval.

Routes each card to the appropriate skill via label-based matching (app/agents/card_routing.py): bug maps to bug-fix, feature maps to add-set, chore maps to quick.

Dispatches agent runs through AgentSessionManager.start_run() with atomic lock_card CAS to prevent double-claiming.

Per-card retry limit: 3 attempts before the card is moved to the Blocked column.

Commit Trailers

Every git commit produced during an autopilot-dispatched run automatically receives two trailers injected by the runtime:

Autopilot-Card-Id: <card_id> -- links the commit to the originating kanban card

Autopilot-Run-Id: <run_id> -- links the commit to the specific agent run

These trailers are injected by a PreToolUse hook (inject_commit_trailers in app/agents/permission_hooks.py) that detects simple git commit Bash commands and appends --trailer flags when the card_id and run_id ContextVars are bound.

Important

This skill is NOT invoked via /rapid:autopilot in the CLI. It runs automatically when the web backend is running and columns have autopilot enabled via the kanban board UI.

Autopilot

The autopilot is a backend-managed background worker, not a CLI-invoked skill. It runs as a lifespan-managed poller inside the RAPID web backend (app/agents/autopilot_worker.py).

How It Works

Polls autopilot-enabled columns for unclaimed kanban cards on a 60-second interval.
Routes each card to the appropriate skill via label-based matching (app/agents/card_routing.py): bug maps to bug-fix, feature maps to add-set, chore maps to quick.
Dispatches agent runs through AgentSessionManager.start_run() with atomic lock_card CAS to prevent double-claiming.
Per-card retry limit: 3 attempts before the card is moved to the Blocked column.

Commit Trailers

Every git commit produced during an autopilot-dispatched run automatically receives two trailers injected by the runtime:

Autopilot-Card-Id: <card_id> -- links the commit to the originating kanban card
Autopilot-Run-Id: <run_id> -- links the commit to the specific agent run

Important

This skill is NOT invoked via /rapid:autopilot in the CLI. It runs automatically when the web backend is running and columns have autopilot enabled via the kanban board UI.