uptime-incident-triage

Uptime Incident Triage Skill

Use this skill when an MCP server or monitored service may be unhealthy.

Workflow

1. Identify the affected registered server and expected transport.
2. Run check_server for the narrowest live signal.
3. Use get_uptime to distinguish a new failure from recurring instability.
4. Review get_dashboard or get_report when multiple targets may be affected.
5. Classify impact as down, degraded, flaky, recovered, or unknown.
6. Report evidence, likely owner action, and the next verification step.

Error Handling

Inspect stable error.code and error.remediation fields. Register missing targets before retrying, and do not enable local stdio unless the target and executable are trusted.

Safety

Health inspection is read-only. Do not modify registrations or alert thresholds without explicit approval.