# Incident: <one-line title>

> Copy this file into `docs/incidents/YYYY-MM-DD-<short-name>.md` the day something goes wrong.
> Write it while it's fresh. Short and true beats long and late.

- **Date:** YYYY-MM-DD
- **Impact:** what visitors or customers saw, and for how long (for example: "site down, about 40 minutes")
- **Status:** fixed / mitigated / still open

## What happened

Two or three sentences a stranger could follow.

## Timeline

All times in one time zone. Note how you found out.

- HH:MM — first sign of trouble (an alert, a customer email, you noticed)
- HH:MM — what you checked first
- HH:MM — the fix went in
- HH:MM — confirmed working

## Cause

The real cause, not the first symptom. If you don't know yet, say so.

## The fix

What you changed, with the commit or setting.

## What I ruled out

List every cause you tested and cleared, and how. This saves the next person (or your AI agent)
from testing them again.

- Ruled out: <cause> — because <evidence>

## Where the lesson goes

Tick each place you put it. A lesson written in only one place gets forgotten.

- [ ] A rule in the agent's instruction file (CLAUDE.md), if it should never happen again.
      One or two sentences, linking back to this note.
- [ ] A memory or notes entry, for facts about the environment that aren't in the code.
- [ ] A test or check, so nobody has to remember. This is the strongest option.
- [ ] Your knowledge notes, so the next big change starts from what's already known.

## Follow-ups

- [ ] <anything still to do, with a date>