← ClaudeAtlas

postmortemlisted

Write a blameless incident postmortem from a timeline, chat log, alerts and notes, with impact, contributing factors, what went well, and specific owned action items. Use when the user asks for a postmortem, incident review, RCA, or incident report.
polytheseus/sre-skills · ★ 0 · DevOps & Infrastructure · score 72
Install: claude install-skill polytheseus/sre-skills
# Blameless postmortem A postmortem exists to make the system safer. It isn't there to record who made a mistake. Write it so that an engineer who wasn't involved can understand what happened and why the actions will help. ## Inputs to gather Ask for, or collect from the repo and terminal, whatever exists: - The incident timeline (from `incident-triage` if it was used), the chat channel export, the pager or alert history - Graphs or numbers for impact: error rate, latency, affected users or requests, revenue - The changes involved: PRs, deploys, config changes - Start, detection, mitigation and resolution times If key facts are missing, write `TODO(owner?): …` in the doc instead of guessing. **Never invent numbers or timestamps.** ## Blameless language - Describe actions and the system, not people's character: write "the deploy pipeline allowed X". Don't write "Bob forgot X". - Use roles, not names, in the narrative ("the on-call engineer"). - Treat a human error as the start of the inquiry, not the conclusion. Ask why the system made that mistake easy, and why it wasn't caught. - Avoid counterfactual blame ("should have", "failed to"). State what happened and what information was available at the time. ## Analysis Look for **contributing factors** rather than a single root cause. Real incidents usually need several things to go wrong together. Use these lenses: 1. **Trigger**: what change or event started it? 2. **Why did it cause harm?** The missing safegua