| name | clawdad-incident-triage |
| description | Use when Clawdad failures, repeated failed messages, Watchtower pauses, session import issues, or delegate stalls need root-cause triage. |
Triage Clawdad incidents from signals to root cause.
Steps:
- Capture the exact failed command, project path, session ID, run ID, lane ID, and timestamp.
- Inspect Clawdad state, delegate status, mailbox status, Watchtower feed, and recent run events.
- Separate transport/session binding failures from delegate semantic failures.
- Check whether a soft Watchtower finding was incorrectly treated as a pause.
- Apply the smallest generalized fix and add regression coverage.
- Verify the fix with doctor commands and a targeted live or test run.