Monitoring Expert guides Claude through making a service observable. It starts from the service level indicators and critical paths, instruments the code, checks that data arrives, then builds dashboards and alerts on top.
The skill carries working examples for Pino structured logging, prom-client counters and histograms, OpenTelemetry spans, a Prometheus alert rule and a k6 load test, plus reference files for profiling and capacity planning.
What You Get
- A five-step workflow from assessing SLIs to shipping alerts, with a check that data arrives before dashboards are built
- Ready examples for Pino JSON logging, prom-client metrics, OpenTelemetry tracing, a Prometheus alerting rule and a k6 ramp-up test
- Reference files for structured logging, Prometheus metrics, OpenTelemetry, alerting rules, dashboards, performance testing, profiling and capacity planning
- A list of rules: structured logs, request IDs, alerts on critical paths, no sensitive data in logs, no alert on every error
Common Use Cases
- Adding request-level metrics and a /metrics scrape endpoint to a Node.js API
- Replacing string-interpolated console logs with structured JSON logs that carry correlation IDs
- Writing a Prometheus alert for a high 5xx error rate and checking it against false positives
- Running a k6 or Artillery load test with latency and error-rate thresholds
- Profiling CPU or memory bottlenecks and forecasting capacity needs
What Makes It Unique
Most observability prompts cover one pillar. This skill covers logs, metrics, traces, dashboards, alerting, load testing, profiling and capacity planning in one workflow, and it puts a verification step in the middle: data must arrive at the collector before dashboards are built, and alerts are checked for false-positive floods before they ship.
Under The Hood
The skill is instructions and reference markdown that Claude loads on demand; it ships no scripts. The examples and references describe commands and endpoints such as an OTLP trace endpoint, a metrics scrape route and a k6 test that sends HTTP requests to a target API. When Claude follows it, it edits application code, configuration and alert rule files in your project, and may run load tests against the URL you give it.