Postmortem Autopilot
Draft a proper postmortem from incident data, not from memory at 2 AM
Pulls together deployment logs, alert timelines, Slack threads, and monitoring data to draft a structured postmortem. You review and add human context — it handles the tedious timeline reconstruction and formatting.
INGREDIENTS
PROMPT
Create a skill called "Postmortem Autopilot". After an incident, help me write a postmortem. Gather data from available sources: - Recent deployments (git log, CI/CD deploy logs) - Alert timeline (monitoring tool exports or screenshots I provide) - Relevant log excerpts - Any Slack threads or notes I paste in From this data, generate a postmortem with these sections: 1. Incident Summary (what happened, impact, duration) 2. Timeline (chronological events with timestamps) 3. Root Cause Analysis (the technical cause and why it wasn't caught) 4. Contributing Factors (process, tooling, or knowledge gaps) 5. Action Items (specific, assignable, with priority: P0/P1/P2) 6. Lessons Learned Use blameless language throughout. Flag where my human input would strengthen the document (decision points, communication gaps, things the data can't capture).
How It Works
After an incident, the last thing anyone wants to do is spend 3-5 hours
reconstructing what happened. This skill gathers the data trail and
produces a first draft: timeline, root cause analysis, contributing factors,
and action items.
What You Get
- Auto-generated incident timeline from deployment logs, alerts, and monitoring data
- Root cause analysis based on the evidence chain
- Contributing factors identification (process gaps, missing monitoring, documentation gaps)
- Action items categorized by priority and assignee
- Formatted in your team's postmortem template (Google SRE, PagerDuty, custom)
- Blameless language throughout
Setup Steps
- After resolving an incident, point your Claw at the relevant data sources
- Provide the incident time window and affected services
- Review the draft, add human context (decisions made, communication gaps)
- Share with the team for review
Tips
- The sooner after the incident you run this, the fresher the data
- It drafts blameless — if it catches itself blaming, it reframes
- Add your own observations about what was confusing or slow during response
- Use the action items as a starting point, then prioritize with the team
- Pairs naturally with Log Detective for deeper root cause analysis