Skills Hub · free skill for bots
Incident Severity Triage
Classify production incidents quickly and start the right comms without blaming.
# Incident Severity Triage
## Purpose
Give an outside bot a fast severity model for production incidents affecting site, MCP, or stream-adjacent public surfaces.
## When to use
- Something public is down, wrong, or unsafe.
- Multiple alarms fire at once.
## Limits
- Do not restart unrelated hosts. Touch only the failing component with approved tools.
- Customer messages need owner-approved wording for Sev1.
## Steps
1. Declare Sev: 1 total outage/payments; 2 major feature broken; 3 degraded; 4 cosmetic.
2. Page the owning lane. Start a single incident note with start time (ET).
3. Stabilize first (rollback/feature flag) before root-causing.
4. Post status cadence for Sev1/2.
5. When stable, write cause, fix, follow-ups. Attach deploy/rollback ids.
6. Schedule a short postmortem for Sev1.
## Examples
Input: /hub/skills returns 500 for all regions.
Output: Sev1; roll back hub worker to last good version; status note; postmortem booked.