postmortem-writelisted

# Postmortem Write ## Purpose Generate a blameless postmortem document for an incident. Used by the `sre` and `incident-responder` agents. The output is a draft the incident commander edits — it enforces the section order, forces UTC timestamps, and requires every action item to have an owner and due date before the doc passes lint. ## Scope - Reads incident artifacts the operator points it at: - chat transcript (Slack export, Discord export, generic `messages.json`) - alert firings within the incident window - deploy log around the incident window - any operator-written notes - Produces a postmortem markdown at `postmortems/YYYY-MM-DD-<slug>.md`. - Enforces the blameless-postmortem template: timeline in UTC, impact quantified, root cause stated, what-worked / what-didn't, action items with owners + due dates. - Does NOT publish to a wiki or send to stakeholders; the operator commits and notifies separately. ## When to use - Within 5 business days of an incident. Memory decays fast; the fresher the writeup, the more useful. - For sev-1 and sev-2 incidents always. Sev-3: optional, but recommended if the incident revealed a process gap. - After a near-miss that didn't page but should have (latent failure mode caught by chance) — call it a "near-miss postmortem" with the same template. ## When NOT to use - For routine deploy rollbacks that worked. A rollback is not an incident; it's the system working as designed. - As performance review mat