| name | incident-postmortem-writer |
| description | Writes a structured incident postmortem (post-incident review) from raw notes,
timelines, and team input. Use this skill whenever a user says "write a postmortem",
"document this incident", "write an incident report", "help us do a blameless
retrospective on this outage", "turn these incident notes into a PIR",
"write a post-incident review", or "document what happened during the outage".
Also activate when someone shares a timeline of events and asks for a structured
incident document. Do NOT use for live incident response, writing runbooks for
future incidents, or general meeting retrospectives.
|
Incident Postmortem Writer
Transform raw incident notes, chat logs, and timelines into a structured,
blameless postmortem that helps teams learn and prevent recurrence.
When to Use
- An incident, outage, or production issue has been resolved and needs documentation
- A team wants to run a blameless post-incident review
- Stakeholders need a formal incident report
- You want to track action items from an incident for accountability
When NOT to Use
- During an active incident (this is retrospective documentation)
- Writing runbooks for handling future incidents (use
workflow-documenter)
- General meeting retrospectives
Workflow
Step 1 — Gather Raw Material
Collect from the user:
- Incident summary: What broke? What was impacted? How long?
- Timeline: Ordered events with timestamps (approximate is fine)
- Root cause: The underlying reason(s) the incident occurred
- Contributing factors: What made it worse or harder to detect?
- Resolution steps: How was it fixed?
- Action items: What will prevent recurrence?
- Team involved: Who detected, investigated, and resolved it?
If a Slack/PagerDuty/OpsGenie thread is available, offer to extract the timeline.
Step 2 — Apply Blameless Framing
Postmortems should:
- Focus on systems and processes, not individuals
- Ask "why did the system allow this?" not "why did [person] do X?"
- Identify multiple contributing factors, not a single root cause
- Avoid "human error" as a root cause — dig deeper to find what made error possible
Step 3 — Produce the Postmortem Document
# Incident Postmortem — [Title]
**Date:** [YYYY-MM-DD]
**Severity:** P[0-3] | **Duration:** [N hours/minutes]
**Status:** [Resolved / Monitoring]
**Authors:** [Names]
---
## Summary
[2–3 sentence summary: what broke, what was impacted, how long, how resolved]
## Impact
- **Users affected:** [N users / % of traffic]
- **Services affected:** [Service names]
- **Duration:** [Start] → [End] ([N] hours total)
- **Business impact:** [Revenue, SLA, customer trust]
## Timeline
| Time (UTC) | Event |
|------------|-------|
| HH:MM | [Event description] |
| HH:MM | [Alert fired / person paged] |
| HH:MM | [Investigation began] |
| HH:MM | [Root cause identified] |
| HH:MM | [Fix deployed] |
| HH:MM | [Incident resolved] |
## Root Cause Analysis
### Primary Cause
[What was the direct technical cause?]
### Contributing Factors
- [Factor 1 — e.g., monitoring gap]
- [Factor 2 — e.g., insufficient testing of edge case]
- [Factor 3 — e.g., manual process that was error-prone]
## What Went Well
- [Thing 1 — e.g., quick detection via alerts]
- [Thing 2 — e.g., team communication was clear]
## What Could Be Improved
- [Improvement 1]
- [Improvement 2]
## Action Items
| Action | Owner | Priority | Due Date |
|--------|-------|----------|----------|
| [Preventive action] | @person | P[0-3] | [Date] |
| [Detection improvement] | @person | P[0-3] | [Date] |
| [Process change] | @person | P[0-3] | [Date] |
## Lessons Learned
[2–3 sentences on key takeaways for the team and organization]
Step 4 — Review for Tone
Check that the document:
- Uses neutral, blameless language throughout
- Does not name individuals as causes (names OK in "who investigated" contexts)
- Is specific enough to be useful for prevention
- Has concrete, assigned action items (not vague "we should improve monitoring")
Output Format
A single Markdown document structured as shown above.
Optionally provide a shortened executive summary (5 sentences) for stakeholder communication.
Safety & Confirmation
- Flag any statements that assign blame to individuals and rephrase before finalizing.
- Never share the postmortem externally (customer-facing) without human review — some details may be sensitive.
- If revenue or customer data numbers are involved, confirm with the user before including exact figures.