Observability Service Health

This is your go-to for checking if an APM service is actually healthy or just pretending to be. It pulls from SLOs, firing alerts, ML anomalies, throughput, latency percentiles, error rates, and dependency health to give you a real picture. The interesting bit is the correlations script, which surfaces attributes like host, pod, or service version that are overrepresented in slow or failing transactions. It uses ES|QL against APM traces and metrics indices by default, with fallbacks to Elasticsearch APIs when needed. Built by Elastic, so it knows where to look in their observability stack. Use it when you need to diagnose whether a service degradation is real, what's causing it, and whether infrastructure like OOM kills or CPU throttling is involved.

Install to Claude Code

npx -y skills add elastic/agent-skills --skill observability-service-health --agent claude-code