What problem does it solve? Teams often cannot answer "is the service broken for users right now" because metrics are incomplete, alerts are noisy or missing, and dashboards bury the signal. This Skill audits the full observability stack so gaps are found before an outage does. ## Core Features & Use Cases - Metrics Coverage Audit: Verifies RED (rate, errors, duration) and USE (utilization, saturation, errors) metrics exist per service, and checks the Prometheus scrape pipeline end-to-end with curl commands against /targets, /metrics, and query APIs. - Alert Quality Review: Applies the actionable/urgent/user-impacting test to every alert, enforces ratio-based expressions with for: durations, severity split between page and warn, and absence-of-data detection via absent(). - Dashboard and Log Inspection: Checks dashboard layout (RED overview, saturation, dependencies, deploy annotations) and log hygiene (structured JSON, request_id propagation, no secrets, retention configured). - Use Case: After deploying a new service, run this review to confirm Prometheus is actually scraping it, error-ratio alerts exist with runbook links, and the on-call dashboard answers health questions in under 30 seconds. ## Quick Start Ask the agent to review the monitoring setup for your service and report critical observability gaps, noisy alerts, and recommended additions.
