dev-team

Spin up the dev team to work on project board tickets. Follow these steps:

Step 0 — Preflight checks (MANDATORY — do this before spawning anything)

Verify all required agent prompt files exist:

for f in \
  .claude/agents/dev/pm-agent.md \
  .claude/agents/dev/team-lead-agent.md \
  .claude/agents/dev/frontend.md \
  .claude/agents/dev/backend.md \
  .claude/agents/dev/test-agent.md \
  .claude/agents/dev/qa-agent.md; do
  [ -f "$f" ] && echo "OK: $f" || echo "MISSING: $f"
done

If ANY file is reported MISSING, stop immediately and output:

PREFLIGHT FAILED: The following required files are missing:
  - {list missing files}
Aborting. Fix the missing files and re-run the skill.

Do NOT proceed to Step 1 until all files are confirmed present.

Step 1 — Create team

Create a team named "dev-team".

Then immediately run:

sleep 3

This gives the teammate snapshot time to initialize before spawning, avoiding the race condition in teammate initialization.

Step 2 — Spawn teammates

Spawn two teammates, BOTH as subagent_type: "general-purpose":

• PM (name: "pm") — read the full contents of .claude/agents/dev/pm-agent.md and use it as the prompt
• Team Lead (name: "dev-lead") — read the full contents of .claude/agents/dev/team-lead-agent.md and use it as the prompt

IMPORTANT: Never use any subagent_type other than "general-purpose" — this applies to both teammates AND any sub-agents they spawn (dev, test, QA). Named agent types like pm-board-manager or team-lead-coordinator do not provide the required tool access.

Step 3 — Verify startup (with retry)

Wait for dev-lead's first message containing its ENV CHECK result (up to 5 minutes).

If dev-lead responds with ENV CHECK OK → team started successfully, proceed to Step 4.

If ANY of the following occur:

• dev-lead responds with FATAL: Agent tool missing
• dev-lead produces an internal error
• dev-lead gives no response within 5 minutes

→ Enter the retry sequence below. Maximum 3 attempts total.

Retry sequence

1. Track attempt number (starts at 1, max 3)
2. Shut down both teammates ("pm" and "dev-lead")
3. Run:

   sleep 5

4. Delete and recreate the team "dev-team"
5. Run:

   sleep 3

6. Re-spawn both teammates exactly as in Step 2
7. Wait again for dev-lead's ENV CHECK response (up to 5 minutes)
8. If ENV CHECK OK → proceed to Step 4
9. If still failing and attempt < 3 → go back to step 1 of retry sequence
10. If still failing after attempt 3 → abort:

    SPAWN FAILED after 3 attempts.
    This is likely the race condition bug in teammate initialization.
    Recommended: wait 2-3 minutes and re-run /dev-team.
    If it keeps failing, restart your claude session and try again.

    Shut down all teammates and delete the team. Stop.

Step 4 — Monitor

Monitor the team while it runs. You are observing only — do not interfere with the PM/team-lead loop unless:

• dev-lead goes silent for more than 10 minutes mid-cycle → send:

  Watchdog ping: You have been idle for 10 minutes. Report your current status.

• dev-lead receives two consecutive watchdog pings with no response → abort:

  WATCHDOG TIMEOUT: dev-lead is unresponsive. Shutting down team.

  Shut down both teammates and delete the team.

Step 5 — Shutdown

When dev-lead produces its Final Report OR the watchdog aborts the run:

1. Shut down both teammates ("pm" and "dev-lead")
2. Delete the team "dev-team"
3. Output a summary:

   DEV TEAM RUN COMPLETE
   Outcome: {completed | aborted | watchdog-timeout | preflight-failed}
   Cycles completed: {N}
   Tickets moved to Review: {list or "none"}
   Tickets flagged: {list or "none"}
   Spawn attempts needed: {N}