Kubernetes Pod Crash Investigator
Diagnoses CrashLoopBackOff and OOMKilled pod failures using the Kubernetes API via kubectl and the official kubernetes-client/python SDK. Correlates container logs, resource limits, and node conditions for root cause analysis.
npx skills add agentskillexchange/skills --skill kubernetes-pod-crash-investigator-3
The Kubernetes Pod Crash Investigator automates the diagnosis of pod failures in Kubernetes clusters. It connects to clusters via the official kubernetes-client/python SDK or kubectl CLI to gather comprehensive failure context.
What this skill actually does
When a pod enters CrashLoopBackOff or OOMKilled state, the skill retrieves container logs from the current and previous instances, examines resource requests and limits, and checks node-level conditions including memory pressure and disk pressure events.
The investigator correlates multiple data sources including Kubernetes Events API, container exit codes, liveness/readiness probe configurations, and PodDisruptionBudget status. It builds a timeline of events leading to the crash and identifies common root causes.
Supported diagnostics include memory leak detection through OOM score analysis, misconfigured health probes, image pull failures with registry authentication issues, and init container dependency failures. Results are formatted as actionable runbook steps with suggested kubectl commands for remediation.