Observability And SLO

Use this skill when a change needs production visibility, alerting, SLOs, or dashboard/runbook updates.

Workflow

1. Identify the user-visible behavior that should be measured.
2. Define SLIs before alerts: availability, latency, correctness, freshness, durability, or throughput.
3. Define SLO targets and error budget policy in docs/harness/SLO_POLICY.md.
4. Map telemetry:

- logs for discrete events and audit trails - metrics for aggregate health and alerting - traces for request paths and dependency latency

1. Add alert criteria only when a human action is expected.
2. Update docs/harness/OBSERVABILITY.md and docs/harness/OPERATIONS_RUNBOOK.md.
3. Add dashboard and runbook links or placeholders when live URLs are not available.

Alert Quality Checklist

• The alert maps to a user impact or fast-moving risk.
• The alert has an owner and runbook.
• The runbook includes first diagnostics, rollback, and escalation.
• The threshold avoids paging for harmless noise.
• The signal can distinguish service failure from dependency failure where possible.

Required Output

Return SLIs, SLOs, telemetry changes, alerts, dashboards, runbook updates, and remaining blind spots.