Incidents

Severity

Level	Means	Example
SEV1	Members are affected right now	Checkout failing, the app down, data exposed
SEV2	Degraded, or SEV1 if nobody acts	A shipped build rejecting a payload, alarms storming
SEV3	Contained, no member impact	A cron failing quietly, a staging box down

Who does what

Anybody may declare an incident, and declaring one too eagerly is never the mistake. The declarer owns communication until they hand it over explicitly. For a SEV1, say something to affected members before it is fully understood; silence reads worse than an unfinished answer.

Postmortems

Every SEV1 and SEV2 gets a postmortem within a week, using templates/postmortem.md.

They are blameless. Written about the system, never about a person. "The deploy skipped the migration" is a finding. "He forgot the migration" is not, and it is also less true: if one person forgetting can break it, the system is what needs fixing.

Index

Nothing yet. Seeded in stage 3.