[Gastown] Polecats fail to start/dispatch — all beads fail with assignee_agent_bead_id null after 5 attempts

[fanioz]

What happened?

Summary

Polecats in the pecutai rig (b202c97f-4cfd-4c43-9b11-f47b410a18b6) consistently fail to start work on dispatched beads. This has happened across TWO consecutive convoys with an identical failure pattern, despite verified repo/git access.

Evidence

Convoy 1: 42221cf4 (closed)

• 3 implementation beads: all failed (5 dispatch attempts each)
• 5 merge_request review beads: all failed (5 dispatch attempts each)

Convoy 2: 1595f70f (closed)

• 3 implementation beads: all failed (5 dispatch attempts each)
• 5 merge_request review beads: all failed (5 dispatch attempts each)

Key anomaly

After 5 dispatch attempts, ALL beads show assignee_agent_bead_id: null. The polecats (Toast, Maple) never hook onto any bead. They remain idle with dispatch_attempts: 0 on the agent records, yet last_activity_at timestamps match the bead dispatch times — suggesting the dispatch fires but the polecat never actually starts working.

What I verified is NOT the problem

1. Repo access: The repo (fanioz/pecutai, private) has a fine-grained PAT with Contents/Metadata/PR/Branches read+write. gh api repos/fanioz/pecutai/contents/ returns data successfully.
2. Git clone: git clone https://github.com/fanioz/pecutai.git succeeds (with gh auth git-credential helper configured).
3. Repo structure: The files referenced in bead descriptions exist — internal/agent/packs.go, internal/action/types.go, internal/action/custom_example.go are all present.
4. Branch creation: No convoy feature branch was ever created on the remote (convoy/.../head returns 404), confirming polecats failed before any git work.

Timeline (convoy 2)

• 12:54:28 — convoy created and started, 3 beads dispatched
• 13:05:21 — beads 1 & 2 failed after 5 dispatch attempts (~11 min)
• 13:17:56 — bead 3 failed after 5 dispatch attempts
• 13:19:49 — first merge_request review bead created, then failed
• 14:21:49 — last merge_request review bead failed

Reproduction

1. Create any bead in rig b202c97f-4cfd-4c43-9b11-f47b410a18b6
2. Dispatch it
3. It will fail 5 times with the polecat never actually starting

Suspected cause

The polecat dispatch/startup mechanism is failing. The beads get dispatch_attempts incremented but the polecat agent never actually begins execution (no in_progress state, no code changes, no branch creation). Possible causes: container startup failure, agent runtime error, or the polecat environment's git credentials not matching the updated PAT.

Town

6fa55c35-46e1-4a92-a478-cb50e99f256d

Area

Agent Dispatch / Scheduling

Context

• Town ID: 6fa55c35-46e1-4a92-a478-cb50e99f256d
• Agent: Mayor (88e1c75e-3d01-4dd7-a61f-6d11f94730f7)
• Rig ID: b202c97f-4cfd-4c43-9b11-f47b410a18b6

Recent Errors

No explicit error messages visible to mayor — beads simply show status "failed" after 5 dispatch attempts with assignee_agent_bead_id: null. Polecats remain idle. No logs accessible to mayor.

Filed automatically by the Mayor via gt_report_bug.