Triggers on 'post-mortem', 'log this failure', 'what went wrong'. Auto-detects failure signals (crashes, rollbacks, regressions, wasted time) and prompts user. Appends entry to main/post-mortems.md.
Post-Mortem System
"post-mortem""postmortem""log this failure""write a post-mortem""what went wrong"AI watches for these signals and prompts the user:
| Signal | Phrase Examples |
|---|---|
| Deployment failure | "it crashed", "pod is failing", "image pull error", "rollback" |
| Test regression | "tests are broken", "was passing before", "something broke" |
| Architecture reversal |
| "undo this", "we need to revert", "this approach doesn't work" |
| Wasted time | "wasted hours", "dead end", "that didn't work at all" |
| Security incident | "exposed secret", "accidentally committed", "vulnerability" |
| Data loss | "data is gone", "migration failed", "backup didn't work" |
On detection, AI asks: "That didn't go as planned. Worth a post-mortem?"
User says yes → AI fills out the format from post-mortem-core.md.
User says no → move on, no log created.
User says "post-mortem" or "log this failure" → AI immediately starts the post-mortem format.
post-mortem-core.md, ask clarifying questions as neededmain/post-mortems.mdWhen starting work in a domain that has a past post-mortem:
main/post-mortems.md for relevant entries