Generate production-grade monitoring configurations: Prometheus alerting/recording rules, Grafana dashboards as code (JSON), AlertManager routing, and scrape configurations. Use this skill whenever the user asks to create monitoring, alerting, dashboards, observability configs, SLOs, or PromQL queries. Also trigger when the user describes services they want to monitor (e.g., "I need alerts for my API and database"), mentions Prometheus, Grafana, AlertManager, ServiceMonitor, or asks for RED/USE method metrics, SLI/SLO definitions, or dashboard templates. Covers Prometheus rules, Grafana dashboard JSON, AlertManager routing trees, Kubernetes ServiceMonitor/PodMonitor CRDs, and Grafana provisioning files. Target audience is advanced DevOps engineers — skip basics, focus on PromQL patterns, alert design, dashboard structure, and operational signal quality.
Generate production-grade monitoring configurations as code. The audience runs Prometheus + Grafana in production — they want well-designed alerts, dashboards, and recording rules, not tutorials on what a metric is.
references/common-pitfalls.md — Read when dealing with cardinality explosions, alert fatigue, histogram bucket design, or Grafana performance issues.Extract from the user's request (infer where possible, ask only when genuinely ambiguous):
Depending on the request, generate one or more of:
PrometheusRule CRDs or standalone .rules.ymlAdd comments explaining: why specific thresholds, why certain aggregation windows, why a particular PromQL approach. Don't explain what PromQL is; explain why this query catches the problem reliably.
apiVersion: monitoring.coreos.com/v1