Observability and Instrumentation tells an agent to write telemetry alongside the feature. It starts by writing the questions an on-call engineer will ask, then picks logs, metrics or traces to answer each one.
The skill covers structured logging with correlation IDs and entry-point fields, RED and USE metrics with bounded labels, OpenTelemetry tracing, symptom-based alerting, minimal runbooks and a step for verifying that the telemetry works.
What You Get
- A seven-step process from on-call questions to telemetry verification
- A table matching logs, metrics and traces to the questions they answer
- TypeScript examples for Express logging, Prometheus histograms and OpenTelemetry setup
- A runbook template, red flags and a verification checklist
Common Use Cases
- Adding logging and metrics to a new service
- Setting up alerting rules
- Reviewing a pull request that adds retries or queues
- Making production incidents easier to diagnose
What Makes It Unique
The skill is firm about cardinality, with a list of what can and cannot be a metric label, and about naming the entry point when a scheduler, a replay endpoint and a CLI write to one log. It limits alerts to two severities and requires each alert to be actionable.
Under The Hood
The skill is a single SKILL.md of instructions with TypeScript examples. It runs no scripts. The code it asks for adds OpenTelemetry and Prometheus client libraries and sends telemetry to a backend you choose, and it tells the agent to test by inducing errors in staging.