Blameless Post-Mortem

Facilitate a blameless post-mortem for $ARGUMENTS (an incident id, or the most recent incident if empty). Blameless means: focus on systems and contributing factors, never on individual faults.

Steps

1. Reconstruct the timeline.
   • From git: git log --since=<window> --pretty=format:'%h %ad %s' --date=short for changes around the incident.
   • From incident data: query stored RCAReports (via the postgres/sqlite MCP or the filesystem store) for the incident id - detection time, consensus outcome, SkepticVerdict, ConfidenceLevel, any HITLRequest.
2. Five why's - drill from symptom to root cause; record each "why" and its evidence.
3. Contributing factors - what made detection/diagnosis/remediation slower (gaps in runbooks, missing tests, alert fatigue, etc.).
4. Action items - concrete, each with an owner and a tracking issue; route code fixes to python-expert, tests to testing-expert, doc/runbook updates to docs-updater, design changes to app-architect.

Template (docs/post-mortems/YYYY-MM-DD-<incident-id>.md)

# Post-Mortem - <incident-id> - <date>
- Severity: | Duration: | Detected by: | Mode:

## Summary
## Impact
## Timeline (UTC)
| Time | Event | Source |

## Root cause (five-whys)
1. Why … → because …

## What went well
## What went wrong / contributing factors
## Action items
| # | Action | Owner | Agent | Tracking |

Output

Print the post-mortem path. Keep tone factual and blameless throughout. Offer to open the action-item issues via the github MCP.