pod-failure-triage

name	pod-failure-triage
description	Triage Kubernetes pods in CrashLoopBackOff, ImagePullBackOff, Pending, OOMKilled, or probe failure states.

Pod Failure Triage

Use this skill when investigating unhealthy pods or workload rollouts.

Workflow

1. List pods in the target namespace with k8s_get_resources.
2. Describe the affected pod with k8s_describe_resource.
3. Read pod logs with k8s_get_pod_logs; check previous logs when restart count is greater than zero.
4. Inspect owner references and describe the owning Deployment, StatefulSet, DaemonSet, Job, or CronJob.
5. Review events in the namespace with k8s_get_events.
6. Separate symptoms from root cause: restarts, failed probes, and unscheduled pods are symptoms; the cause is the failing dependency, bad image, bad configuration, resource pressure, or application error.

Output

Return:

• Pod, namespace, node, and owner
• Current phase and container waiting or termination reason
• Relevant log or event evidence
• Root cause
• GitOps-safe remediation