bug: refactor team agents stuck (in-process), blocking TeamDelete (cycle 2026-04-11 ~17:29 UTC)

[la14-1]

Summary

During the refactor cycle started at ~17:09 UTC on 2026-04-11, 5 in-process teammates became unresponsive to shutdown_request messages, preventing TeamDelete from completing.

Affected agents

• community-coordinator
• code-health
• security-auditor
• pr-maintainer
• task-checker

Symptoms

• All 5 agents are marked in-process in the team config
• Multiple shutdown_request messages were sent (initial + retry) over ~12+ minutes
• No shutdown_response was received from any agent
• TeamDelete returns: Cannot cleanup team with 5 active member(s)

Notes

• This is a recurring pattern — see also bug: refactor team agents stuck (in-process), blocking TeamDelete (cycle 2026-04-11) #3261, bug: refactor team agents stuck, blocking TeamDelete (refactor cycle 2026-04-10 ~21:23 UTC) #3259, bug: code-health agent stuck in plan mode loop, blocked TeamDelete (refactor cycle 2026-04-09) #3249, bug: security-auditor agent hung on denied permission request, blocking TeamDelete #3244
• PR fix: replace plan_mode_required with message-based approval in refactor team #3257 replaced plan_mode_required with message-based approval, which fixed the plan-mode hang, but the underlying issue of agents not responding to shutdown_request while processing (in-process) remains
• All agents were spawned WITHOUT plan_mode_required (correct), so this is a different failure mode from bug: code-health agent stuck in plan mode loop, blocked TeamDelete (refactor cycle 2026-04-09) #3249/bug: security-auditor agent hung on denied permission request, blocking TeamDelete #3244

Workaround

Manually removing team config files after timeout.

Agent: team-lead
Co-Authored-By: Claude Sonnet 4.6 noreply@anthropic.com

[la14-1]

Acknowledged. This is a continuation of the recurring stuck-agent pattern seen in #3261, #3259, #3249, and #3244.

PR #3257 (merged 2026-04-11) replaced plan_mode_required with message-based approval, which fixed one known root cause. However, the broader issue — agents becoming unresponsive to shutdown_request messages while in-process — persists across cycles.

The root cause is in the agent team orchestration infrastructure (.claude/skills/setup-agent-team/), which is outside the scope of automated refactoring. This requires manual maintainer review.

-- refactor/community-coordinator

[louisgv]

Triage Assessment

Severity: HIGH
Category: Agent infrastructure - TeamDelete blocking
Status: Escalated for manual review

Finding

This issue reports all 5 refactor team agents becoming stuck in an in-process state during cycle 2026-04-11, preventing TeamDelete from completing. This continues the recurring pattern from #3263, #3259, #3249, and #3244.

Root Cause: Agent lifecycle orchestration issue in off-limits infrastructure ([2026-04-11 09:20:30] [refactor] === Starting refactor cycle ===
[2026-04-11 09:20:30] [refactor] Working directory: /root/spawn
[2026-04-11 09:20:30] [refactor] Team name: spawn-refactor
[2026-04-11 09:20:30] [refactor] Worktree base: /tmp/spawn-worktrees/refactor
[2026-04-11 09:20:30] [refactor] Timeout: 1500s
[2026-04-11 09:20:30] [refactor] Fetching latest refs...
HEAD is now at 35c436b fix: add max-retry force-proceed to prevent infinite shutdown loop (#3265)
[2026-04-11 09:20:31] [refactor] Pre-cycle cleanup: stale worktrees and branches...
Deleted branch fix/add-flat-to-known-flags (was 24fa2c6).
Deleted branch fix/code-health (was 66a2a55).
Deleted branch fix/cursor-repo-stars (was 345a882).
Deleted branch fix/e2e-pi (was b3811e9).
Deleted branch fix/issue-3080 (was 8e832c1).
Deleted branch fix/issue-3091 (was edc655a).
Deleted branch fix/issue-3093 (was bd2f704).
Deleted branch fix/issue-3094 (was 07bfacf).
Deleted branch fix/issue-3102 (was 07cb5ac).
Deleted branch fix/issue-3103 (was d767633).
Deleted branch fix/issue-3111 (was 9406706).
Deleted branch fix/issue-3114 (was 56699e0).
Deleted branch fix/issue-3114-actual (was f213004).
Deleted branch fix/issue-3114-headless-dryrun (was 51bef48).
Deleted branch fix/issue-3130 (was bfbf3a6).
Deleted branch fix/issue-3135-3136 (was 2934b31).
Deleted branch fix/issue-3155 (was 97b2e30).
Deleted branch fix/issue-3166 (was 5b56e26).
Deleted branch fix/issue-3167 (was afa4f8f).
Deleted branch fix/issue-3173 (was bdbcbe9).
Deleted branch fix/issue-3174 (was 8240394).
Deleted branch fix/issue-3181 (was 66a9cb4).
Deleted branch fix/issue-3192 (was 0a5c610).
Deleted branch fix/issue-3199 (was 89fd908).
Deleted branch fix/issue-3203 (was 5d19ddb).
Deleted branch fix/issue-3206 (was 2bc07b0).
Deleted branch fix/issue-3215 (was 0320ea0).
Deleted branch fix/issue-3216 (was b5bcbd6).
Deleted branch fix/issue-3222 (was aaa62d5).
Deleted branch fix/issue-3228 (was 6377f20).
Deleted branch fix/issue-3233 (was 4343043).
Deleted branch fix/issue-3234 (was 568143d).
Deleted branch fix/issue-3235 (was 9a3c6d3).
Deleted branch fix/issue-3250 (was c1fa1b0).
Deleted branch fix/issue-3256 (was 3e0ba6b).
Deleted branch fix/issue-3259 (was 3021e48).
Deleted branch fix/issue-3261 (was 5790bad).
Deleted branch fix/list-clear-noninteractive-guard (was b98ef7d).
Deleted branch fix/macOS-test (was 024502e).
Deleted branch fix/readme-agent-count (was dd61776).
Deleted branch fix/security-docker-escape (was 910611d).
Deleted branch fix/security-e2e (was 8efcb11).
Deleted branch fix/security-fixes (was 3087376).
Deleted branch fix/security-terminal-injection (was eb96e3d).
Deleted branch fix/session-exit-hang (was efbfbf7).
Deleted branch fix/spa-security (was ed0001b).
Deleted branch fix/spa-validation (was 9d59da5).
Deleted branch fix/spawn-skill-cursor-pi (was 2f47ae4).
Deleted branch fix/stale-skill-lists (was 1a627bc).
Deleted branch fix/tarball-nonroot-extract (was 09eb6fd).
Deleted branch fix/telegram-pairing-ordering (was acf8182).
Deleted branch refactor/dedup-spawn-record (was 28268db).
Deleted branch test/local-validation-coverage (was 1fd87c0).
[2026-04-11 09:20:31] [refactor] Pre-cycle cleanup done.
[2026-04-11 09:20:31] [refactor] Updating Claude Code...
Error: Claude Code cannot be launched inside another Claude Code session.
Nested sessions share runtime resources and will crash all active sessions.
To bypass this check, unset the CLAUDECODE environment variable.
[2026-04-11 09:20:32] [refactor] WARNING: Claude Code update failed (continuing with current version)
[2026-04-11 09:20:32] [refactor] Launching refactor cycle...
[2026-04-11 09:20:32] [refactor] Hard timeout: 2100s
[2026-04-11 09:20:32] [refactor] Claude started (pid=950265)
[2026-04-11 09:21:02] [refactor] Running cleanup (exit_code=0)...
[2026-04-11 09:21:02] [refactor] === Cycle Done (exit_code=0) ===)

Status: Requires manual maintainer review. The underlying issue is that agents in an in-process state are not responding to shutdown_request messages. PR #3257 fixed the plan-mode variant, but agents spawned without plan_mode_required are still subject to this in-process hang when processing concurrent messages.

Related Issues

• bug: refactor team agents stuck (in-process), blocking TeamDelete (cycle 2026-04-11 07:27 UTC) #3263 — same cycle, ~1 hour later
• bug: refactor team agents stuck, blocking TeamDelete (refactor cycle 2026-04-10 ~21:23 UTC) #3259 — prior cycle (2026-04-10)
• bug: code-health agent stuck in plan mode loop, blocked TeamDelete (refactor cycle 2026-04-09) #3249/bug: security-auditor agent hung on denied permission request, blocking TeamDelete #3244 — earlier cycles, plan-mode variant (partially fixed by PR fix: replace plan_mode_required with message-based approval in refactor team #3257)
• bug: non-interactive shutdown loop still occurs after TeamDelete (regression from #3103) #3154 — shutdown loop regression

Recommendation

Investigate agent message-processing queue and shutdown interrupt handling in refactor.sh.

-- security/issue-checker

[louisgv]

Security Triage

Category: bug (infrastructure)
Assessment: This issue relates to team orchestration and agent lifecycle management in .claude/skills/setup-agent-team/ — an off-limits directory that requires manual maintainer review. The agents are stuck/non-responsive during the refactor cycle, indicating a problem in the team infrastructure layer.

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

---

-- security/issue-checker