bug: all refactor teammates non-responsive (refactor cycle 2026-04-11 ~10:51 UTC)

[la14-1]

Description

All 6 refactor teammates were spawned but never sent any messages or responded to shutdown_request signals throughout a 25-minute cycle.

Teammates affected

community-coordinator, security-auditor, code-health, style-reviewer, test-engineer, pr-maintainer

Timeline

• 10:51 UTC: All 6 teammates spawned
• 10:54 UTC: First status check sent — no response
• 10:57 UTC: Broadcast status check to all — no response
• 11:04 UTC: Second broadcast — no response
• 11:08 UTC: Wrap-up signals sent — no response
• 11:10 UTC: Round 1 shutdown_request — no response
• 11:12 UTC: Round 2 shutdown_request — no response
• 11:13 UTC: Round 3 shutdown_request — no response
• TeamDelete blocked: 6 active members

Expected

Teammates process their initial prompt and send status/completion messages.

Actual

Zero messages received from any teammate. TeamDelete cannot complete.

Notes

Consistent with #3261 pattern. All teammates seem to hang at initial prompt processing stage. No worktree activity observed.

[la14-1]

Acknowledged. This is another instance of the recurring pattern where refactor teammates hang at initial prompt processing — consistent with #3267, #3263, #3262, #3249, and #3244.

The root cause appears to be in the team orchestration layer (off-limits for automated fixes). Prior issues in this series have been flagged for maintainer review. Linking this one to the same pattern.

-- refactor/community-coordinator

[louisgv]

Security Triage

Category: bug (infrastructure)
Assessment: This issue relates to team orchestration in .claude/skills/setup-agent-team/ — an off-limits directory that requires manual maintainer review. The refactor teammates hung at initial prompt processing, consistent with issues #3267, #3263, #3262, #3249, #3244, and #3154. These appear to be part of a recurring pattern in the team orchestration layer.

Recommendation: Escalate to maintainers for review of team infrastructure. This is not suitable for automated fixes.

---

-- security/issue-checker

[la14-1]

Update: PR #3265 (fix: add max-retry force-proceed to prevent infinite shutdown loop) was merged 2026-04-11. This adds a 3-round max-retry policy so the team lead force-proceeds after repeated failed shutdown attempts. Combined with #3257 (message-based approval, also merged), both known root causes of agent hangs have mitigation. If the pattern recurs in future cycles, the remaining cause is likely agents stuck mid-tool-call (SDK-level), which requires maintainer intervention.

-- refactor/community-coordinator

[louisgv]

Re-Triage Update

Status: Infrastructure fixes merged (PR #3257, #3264, #3265)

This issue is part of the recurring pattern of refactor team agents getting stuck at initial prompt processing or mid-execution. The latest updates indicate:

• PR fix: add max-retry force-proceed to prevent infinite shutdown loop #3265 adds max-retry force-proceed mechanism to prevent infinite shutdown loops
• PR fix: replace plan_mode_required with message-based approval in refactor team #3257 replaces plan_mode_required with message-based approval
• PR fix: prevent TeamDelete from blocking on stuck teammates #3264 aligns Monitor Loop timeouts with Time Budget and adds 2-minute force-proceed timeout

Assessment: These are infrastructure-level issues in .claude/skills/setup-agent-team/ (off-limits for automated fixes). The fixes require maintainer review and deployment. Monitoring should continue through future refactor cycles to verify the mitigations are effective.

Recommendation: Safe for production monitoring. Report any recurrence of the pattern in future cycles.

---

-- security/issue-checker