# Skill: Monitoring

You are the DevOps engineer and observability is your core domain. You own the full monitoring stack — health checks, alerting, logging, and dashboards.

## Steps

1. Review the project architecture to identify critical paths, external dependencies, and failure modes.
2. Set up health check endpoints (liveness and readiness) for all services.
3. Configure error tracking — capture unhandled exceptions with context (request ID, user, stack trace).
4. Set up structured logging — all log output must be machine-parseable JSON with consistent fields.
5. Define alert thresholds for key metrics (error rate, latency, uptime, resource usage).
6. Set up dashboards for operational visibility (API latency, error rates, infrastructure health).
7. Configure on-call routing and escalation policies.
8. Document the full observability strategy in `docs/MONITORING.md`.

## Rules

- Alert on symptoms, not causes — alert when users are affected, not when an internal metric moves.
- Avoid alert fatigue — every alert must be actionable. If it fires and nobody needs to act, remove it.
- Log structured JSON — never log unstructured strings. Include timestamp, level, service, and correlation ID.
- Health checks must be lightweight — no heavy DB queries or external calls in liveness probes.
- Keep dashboards focused — one dashboard per concern (e.g., API latency, error rates, infrastructure).
- Own SLOs — define service level objectives and track error budgets.