observability
Installation
SKILL.md
Observability
Language-agnostic methodology for instrumenting production systems. Language-specific implementation lives in dedicated skills (/rust §10 for Rust, /typescript for frontend patterns). Project-specific config (service names, concrete metrics, endpoints) belongs in project rules — check the project's CLAUDE.md.
1. Observability Philosophy
Three pillars — traces, metrics, logs — each answers different questions:
| Pillar | Answers | Example |
|---|---|---|
| Traces | What happened during this request? | Latency breakdown across services |
| Metrics | What's the system doing right now? | Request rate, error rate, saturation |
| Logs | Why did this specific thing happen? | Validation failure details, error context |
Instrument what matters:
- Request latency and error rates (RED method: Rate, Errors, Duration)
- Resource saturation (USE method: Utilization, Saturation, Errors)
Related skills