Observability

observability

Observability skills for designing tracing strategies, instrumenting existing code with structured log points, and analyzing trace logs to diagnose production issues.

Skills

/trace-plan

Design a logging and distributed tracing strategy for a system or agent pipeline. Produces a span tree, structured log schema, log-level policy, and a prioritized list of what to instrument first. Works with any language or framework.

/instrument-code

Add structured log points and trace spans to existing code — minimally and additively, without restructuring logic. Covers structlog, OpenTelemetry, pino, and other popular libraries. Includes guidance on avoiding PII leakage and keeping tests green.

/analyze-traces

Parse structured trace or log data and produce a diagnostic report covering latency hotspots, critical path analysis, error cascades, concurrency issues, and ranked recommendations. Works with OTLP JSON, newline-delimited JSON logs, and plain text.

When to use

Skill	Trigger
/trace-plan	Starting a new project or adding observability from scratch
/instrument-code	Adding log points and spans to specific existing functions
/analyze-traces	Diagnosing a production issue from captured logs or traces

Workflow

/trace-plan  -->  /instrument-code  -->  /analyze-traces
  (design)          (implement)           (diagnose)

Dependencies

• Python: structlog, opentelemetry-sdk, opentelemetry-exporter-otlp
• TypeScript: pino, @opentelemetry/sdk-node

All dependencies are optional — skills adapt to whatever logging library the project already uses.

Sources

Derived from:

• observability/ patterns (anthropic-cookbook)
• OpenTelemetry semantic conventions