What problem does it solve? Go services shipped without observability are black boxes in production: no way to diagnose latency spikes, error bursts, goroutine leaks, or slow dependencies. This Skill guides an AI agent to instrument Go code with the five production signals — logs, metrics, traces, profiles, and RUM — following proven conventions and avoiding common pitfalls like high-cardinality labels and missing trace context. ## Core Features & Use Cases - Structured logging with slog: JSON handlers for production, context-aware log calls for trace correlation, and incremental migration from zap, logrus, or zerolog via bridge handlers. - Prometheus metrics and alerting: correct metric types (Histogram over Summary), naming conventions, PromQL-as-comments above declarations, multi-window burn-rate SLO alerts, and Go runtime alerts for goroutine leaks and GC pressure. - OpenTelemetry tracing and profiling: span placement on service methods, DB queries, and external calls, context propagation, exemplars linking metrics to traces, and pprof/Pyroscope profiling toggled by environment variables. - Use Case: When adding a new HTTP endpoint to a Go service, ask the agent to make it observable — it will declare latency and error metrics with PromQL comments, add spans with error recording, emit structured context-aware logs, and wire alert rules. ## Quick Start Ask the agent to add production observability to your Go service, for example: instrument my HTTP handlers with Prometheus metrics, OpenTelemetry spans, and slog structured logging.
