# <Symptom as the user or alarm reports it>
Services: <names from ci/service-map.json> · Last seen: <date, PR/incident link> · Owner: <name>

## Symptom
<What people see: status code, UI message, alarm name.>

## Confirm it is this
<One or two checks that tell this failure apart from look-alikes. Exact commands or console clicks.>

## Where to look
- CloudWatch `/ecs/<tier>/<service>`; search for: `<key message>`
- <dashboards, ECS events, Kong logs>

## Fix
<Immediate mitigation first (rollback, restart, config flip), then the real fix.>

## Rollback
<Usually: docs/runbooks/ecs-rollback.md>

## Prevent
<Guard, eval case, or CLAUDE.md line added so it doesn't recur.>