Turn a failing pipeline into a fixed build and, when it matters, a blameless postmortem. This workflow triages the CI failure, hardens the workflow so it does not recur, and drafts the incident write-up from your evidence.
Classify the failing run against a taxonomy of GitHub Actions failure modes - permissions and GITHUB_TOKEN scope, missing secrets, matrix expansion, cache misses, trigger mismatches - and pinpoint the offending step with the exact YAML fix.
Review the workflow for the security and speed issues that cause recurring failures: unpinned actions, over-broad permissions, missing concurrency cancellation, no dependency caching, and missing timeouts. Get a severity-ranked findings table with fixes.
When the incident had real impact, turn your timeline, logs, and alerts into a blameless postmortem: summary, impact, UTC timeline, contributing factors, and owner-assigned action items - written to a file your team can reference.
Paste the failing job log alongside the workflow file. The triager correlates the exact failing step to the config line that caused it.
Run /workflow-review even after the fix. Most CI failures trace back to the same handful of issues - unpinned actions, over-broad permissions, no caching - and hardening the workflow stops the next one.
Only write a postmortem when the incident had real impact. When you do, keep it blameless: the /postmortem command focuses on systems and contributing factors, never individuals.