Code
Write a postmortem
A timeline, a root cause separated from the contributing factors, and actions with owners.
Fill it in
Logs, deploy records and notes from the incident.
What broke, and for how long.
Describe systems and decisions, never people.
Your prompt
Write a postmortem for: [the incident]. Build the timeline from timestamps before writing a word of analysis. Everyone's memory of an incident is wrong in the same direction: it feels shorter, and the moment someone first suspected the cause moves earlier than it happened. Separate the ROOT CAUSE from the contributing factors, and be strict about it. Most postmortems list six causes, which means the reader cannot tell which one removing would have prevented this. Describe systems and decisions, never people. "An engineer deployed without running the migration" is a sentence about a person; "a deploy could reach production with an unapplied migration" is one about a system, and only the second has a fix. Actions with an owner and a date. Anything without both is a wish, and a postmortem full of them is why the same incident happens twice. Include what went RIGHT. The detection that worked is worth keeping.
Continue and pick your folderOpens with everything above already filled in.
Why this works
Everyone's memory of an incident is wrong in the same direction: it feels shorter, and the moment someone first suspected the cause moves earlier than it happened. This builds the timeline from timestamps first, separates root cause from contributing factors strictly, and gives every action an owner and a date.