SkillScan: pass
Blameless incident postmortem
Blameless postmortem template with timeline anchors, TTD/TTM metrics, why-chains, owned action items, and a responder interview guide.
Add a trust badge to your README
[](https://waxmark.app/items/incident-postmortem/scan)
---
name: incident-postmortem
description: Produces a blameless incident postmortem - timeline, impact, contributing factors, detection and response gaps, action items with owners - plus the interview questions to gather facts from responders. Use when the user mentions a postmortem, incident review, RCA, root cause analysis, outage write-up, or asks what to ask the on-call team after an incident.
license: CC0-1.0
metadata:
author: Waxmark
version: "1.0.0"
---
# Incident Postmortem
Blameless means the write-up explains why reasonable people made the decisions they made given what they knew. Name systems, signals, and gaps; never name a person as a cause.
## Procedure
1. **Collect raw facts** before interpreting. Ask for or gather: alert timestamps, chat log of the incident channel, deploy history (`git log --since=... --until=...`), dashboards, the status page history, the ticket. Convert everything to one timezone (state it) and ISO-8601.
2. **Interview** each responder separately using `references/interview-questions.md`. Record answers as quotes with time references, not as conclusions.
3. **Build the timeline** first. Every row is `time | source | what happened or was observed | who/what acted`. Mark the four anchors: first bad change, first customer impact, detection, mitigation.
4. **Derive the metrics** from the anchors: time to detect (impact -> detection), time to mitigate (detection -> mitigation), total impact duration.
5. **Identify contributing factors** with "why" chains, at least three deep, until you reach a system or process property rather than a human action. Several factors are normal; a single root cause is a warning sign of a shallow analysis.
6. **Write action items** that are specific, owned, and dated. "Add alert on `queue_depth > 10k` for 5 min (owner: platform, due: 2026-11-01)" passes. "Be more careful with config changes" does not.
7. Fill the template below. Read it once as the on-call engineer; delete anything that reads as blame.
## Template
```markdown
# Postmortem: <short title> - <YYYY-MM-DD>
**Severity:** SEV<1-4> **Status:** Draft / Reviewed / Final
**Owner:** <team> **Reviewers:** <names or roles>
## Summary
<3-5 sentences: what broke, who was affected, for how long, how it was fixed.>
## Impact
- Users affected: <count or %> for <duration>
- Requests failed / data lost / revenue: <numbers, or "none">
- SLO budget consumed: <value>
## Timeline (all times UTC)
| Time | Source | Event |
|---|---|---|
| 14:02 | deploy bot | v2.31.0 rolled to 100% |
| 14:09 | first impact | 5xx rate crosses 2% (not alerted) |
| 14:31 | pager | SLO burn-rate alert fires |
| 14:36 | responder | rollback started |
| 14:41 | mitigated | 5xx back to baseline |
TTD: 22m TTM: 10m Total impact: 32m
## What went wrong
<Mechanism of the failure, in technical detail. Diagrams welcome.>
## Contributing factors
1. <Factor> - because <why> - because <why> - because <system property>
2. ...
## What went well
- <detection, tooling, or communication that worked>
## Where we got lucky
- <things that would have made it worse and did not happen>
## Action items
| # | Action | Type | Owner | Due | Ticket |
|---|---|---|---|---|---|
| 1 | Alert on 5xx > 1% for 5m at 50% rollout | detect | platform | 2026-11-01 | OPS-412 |
| 2 | Canary stage before 100% | prevent | deploy | 2026-11-15 | OPS-413 |
## Open questions
- <facts still unverified>
```
Action item types: `prevent`, `detect`, `mitigate`, `process`. Aim for at least one `detect` item whenever TTD exceeded a few minutes.
## Output
Deliver the completed template as a Markdown file at `{{POSTMORTEM_PATH}}` (default `docs/postmortems/<YYYY-MM-DD>-<slug>.md`) and list the open questions in the reply.