Nothing to hand? Load the — a messy SEV1 log full of blame-language to neutralize — or the , an incident that never reached a customer and is all the more worth writing up.
Paste the notes
Slack scrollback, the on-call log, pager timestamps, a hand-typed recollection — messy is fine. The instant prescan reads it for free while you type and lists what it mechanically found: every timestamp, every blame-flavored phrase ("should have", "human error", someone's "fault"), and the incident signals — deploys, rollbacks, alerts, saturation, data loss, monitoring gaps.
The AI writes it blameless
An honest severity call and classification (including near-miss), a reconstructed timeline keeping your own timestamps and tagged detect / escalate / mitigate / resolve, a 5 Whys chain where every answer carries evidence from your notes and stops at a systemic cause, contributing factors, and what went well and poorly. Every prescan hit is placed or explicitly set aside — every blame phrase is neutralized into a condition the system allowed.
File the actions and go
Prioritized action items — p0/p1/p2, typed prevent / detect / mitigate / process, each ticket-ready and owned by a role. Download them as a plain CSV or as a Jira/Linear-shaped ticket import, copy the whole postmortem for Confluence, Notion or Google Docs (real tables, not pipe characters), copy the short version for #incidents in Slack formatting, or print it for the review meeting. Tick action items off as they land. History is saved to your account, not just this browser, so the post-meeting re-run can be diffed against the draft and the next incident's writeup can reuse this one's root cause and still-open actions.
Derived from the @wshobson/postmortem-writing skill (MIT license).