Kubernetes Incident Runbook
Executes structured incident response procedures for Kubernetes clusters using kubectl, kube-state-metrics, and the Kubernetes Events API. Automates pod crash diagnosis, OOMKill analysis, and node pressure triage.
Executes structured incident response procedures for Kubernetes clusters using kubectl, kube-state-metrics, and the Kubernetes Events API. Automates pod crash diagnosis, OOMKill analysis, and node pressure triage.
npx skills add agentskillexchange/skills --skill kubernetes-incident-runbook
The Kubernetes Incident Runbook skill provides automated incident response procedures for Kubernetes cluster issues. It uses kubectl commands, the Kubernetes API, and kube-state-metrics to systematically diagnose common failure modes including CrashLoopBackOff, OOMKilled, ImagePullBackOff, and node NotReady conditions.
When triggered, the skill follows a structured diagnostic tree. For pod failures, it inspects container exit codes, retrieves previous container logs via kubectl logs –previous, checks resource requests/limits against actual usage from metrics-server, and examines events for scheduling or volume mount failures.
For node-level issues, it analyzes node conditions (MemoryPressure, DiskPressure, PIDPressure), checks kubelet logs, inspects systemd service status, and correlates with cloud provider instance health. The skill understands taints, tolerations, and affinity rules that may cause scheduling failures.
Advanced capabilities include tracing network connectivity issues using CoreDNS logs and NetworkPolicy analysis, diagnosing PersistentVolumeClaim binding failures across storage classes, and identifying resource quota exhaustion across namespaces. All findings are compiled into structured incident reports with remediation steps.