What problem does it solve? Incident write-ups often devolve into blame, vague timelines, and action items nobody verifies. This Skill produces a disciplined, blameless postmortem grounded in evidence, so the document actually prevents recurrence instead of gathering dust. ## Core Features & Use Cases - Evidence-based timeline reconstruction: Every timeline entry cites an alert, deploy log, dashboard, or chat timestamp, with memory-based entries explicitly labeled. - Trigger vs root cause vs symptom separation: Pushes past the first plausible cause to the missing safeguard that made the trigger sufficient. - Repeat-class detection: Scans prior postmortems for the same subsystem and surfaces recurrences prominently in the Summary. - Owned, dated, verifiable action items: Classifies items as prevention, mitigation, or detection, and stubs runbooks in guides/ for guide-shaped items. - Use Case: After a 47-minute checkout outage caused by a config change, ask the agent to write the postmortem — it gathers evidence, builds a sourced timeline, identifies the 16-minute detection gap, links prior pool-exhaustion incidents, and files the document at postmortems/2024-03-02-payments-pool-exhaustion.md. ## Quick Start Ask the agent to write a blameless postmortem for yesterday's checkout outage using the write-a-postmortem skill, including the timeline, root cause, and action items.
