Stop the bleeding first, explain later — no negotiation.
Severity by criteria, one commander per incident, mitigation before diagnosis, communication on a fixed cadence and blameless postmortems whose actions become gates, not promises to be more careful.
$ schematize install incidentNon-negotiable floors
Never switched off "to go faster". Full list in the skill.
Severity by objective criteria
Not by mood: when in doubt go up a level, and reclassifying is stated and recorded.
One IC per incident
In SEV1/2 the IC does not touch the system; handover of command is explicit.
Mitigate before diagnosing
Stop the bleeding first (revert the last change), one lever at a time, all narrated in the channel.
Irreversible action only by IC decision
Destructive and irreversible steps need a recorded decision and a rollback thought out beforehand.
Silence is a failure
SEV1/2 has a first note and a fixed cadence, always ending with the time of the next; the status page lives outside the affected infra.
Exposed personal data is SEV1
Until proven otherwise, and it starts the LGPD/ANPD clock immediately; the decision to notify or not is written with its reasoning.
Blameless postmortem on SEV1/2
Timeline, quantified impact and a chain of whys; no person names and no "human error".
Actions have an owner and a date, machine over attention
Gates, tests and alerts beat vigilance; a recurring incident becomes an ADR + floor + CI gate.
Commands
Installed in .claude/commands/. Run /inc-help anytime.
/inc-ccContext compact: writes the handoff to the archive and runs /compact./inc-claudeCreates or merges the always-on CLAUDE.md for the incident response baseline at the repo root (won't overwrite other skills' blocks)./inc-commsDrafts the status update (status page, customer, internal) at the cadence of the severity./inc-declareDeclares the incident: classifies SEV from the table, names IC/comms/ops/scribe, opens the channel and timeline./inc-handoffWrites the context handoff (context.md + checklist.md) to the archive, without compacting./inc-helpLists every command and what it does./inc-loadForce-loads the whole incident response normative body and applies it for the session./inc-lgpdAssesses the personal-data trigger, gathers the facts and records the decision on the ANPD and data subjects./inc-postmortemBuilds the blameless postmortem from the timeline and runs the check-postmortem.sh gate./inc-reviewReviews a postmortem and its action items: runs the gate, then reads what the machine cannot.For AI agents
Install with no human in the loop. Metadata in llms.txt.
- Run: git clone https://github.com/schematizeme/skill-incident.git && ./skill-incident/install.sh
- Or download: https://github.com/schematizeme/skill-incident/releases/latest/download/skill-incident.zip and unzip into .claude/skills/
- Copy assets/CLAUDE.md to the repo root (always-on rule).
- Confirm with /inc-help.
Thảo luận
Đang tải…