SkillHub

team-pilot

v0.1.1

Set up and operate TeamPilot from zero on a clean machine. Use when cloning the TeamPilot repository, installing dependencies, starting the multi-agent runtime, driving missions in auto/manual mode, replaying runs in web UI, and troubleshooting first-run issues.

Sourced from ClawHub, Authored by Tiankuo Yao

Installation

Please help me install the skill `team-pilot` from SkillHub official store. npx skills add moolean/team-pilot

Bootstrap and run TeamPilot end-to-end.

Zero-to-run setup (clean environment)

  1. Ensure prerequisites:
  2. git --version
  3. node --version (recommend Node 20+)
  4. npm --version
  5. Clone repository:
  6. git clone <TEAM_PILOT_REPO_URL> team-pilot
  7. cd team-pilot
  8. Install dependencies:
  9. npm install
  10. Start service:
  11. npm run up
  12. Open web UI:
  13. http://localhost:3333

If EADDRINUSE appears, stop the process occupying port 3333 or start with another port: - PORT=3334 npm run up - then open http://localhost:3334

  1. In Mission Control, select a template.
  2. Set execution mode:
  3. auto for smoke tests
  4. manual for deterministic demos and controlled outputs
  5. Fill mission goal and launch.
  6. Confirm graph starts changing (agent lane → task lane → result lane).

Manual mode operation standard

For each task, update in two steps:

  1. task_started (status=running, progress around 20-60)
  2. task_done (status=done, progress=100, complete output)

Endpoint: - POST /api/tasks/:taskId/update

Suggested payload fields: - status, progress, output, score, eventType, assignee

Replay workflow

  1. Switch to Replay mode.
  2. Load a run with frames > 1.
  3. Scrub slider to inspect each frame.
  4. Open node detail to verify input/output at each step.

Minimal API checklist

  • POST /api/missions (create mission)
  • GET /api/state (live state)
  • POST /api/tasks/:taskId/update (manual updates)
  • GET /api/runs (list traces)
  • GET /api/runs/:missionId (load replay trace)
  • POST /api/missions/:missionId/mode (auto/manual switch)

Troubleshooting

  • UI opens but no updates:
  • verify server log has no crash
  • verify WebSocket /ws is connected
  • Replay seems flat:
  • ensure selected run has frames > 1
  • ensure task updates were posted to /api/tasks/:taskId/update
  • Mission stuck queued:
  • check dependency chain and assignee spelling
  • in manual mode, ensure leader is sending task updates

Output quality bar (content pipelines)

  • PLAN: goal, structure, acceptance criteria
  • RESEARCH: sources, extracted facts, citations
  • VERIFY: confidence, risks, uncertainty
  • BUILD: complete final draft
  • DELIVER: publish-ready summary + readiness note