K8s Debug

k8s-debug

Kubernetes cluster debugging skill for Claude Code. Guides safe, structured investigation of pod failures, resource pressure, network issues, and cluster health using local kubeconfigs.

Skill

k8s-debug:k8s-debug

Triggered automatically when you describe a Kubernetes problem — crashing pods, OOMKilled containers, pending resources, unreachable services, node pressure, or anything kubectl-related.

Trigger examples:

• "why is my pod in CrashLoopBackOff?"
• "debug the staging namespace"
• "pod not starting in production"
• "check cluster health"
• "service not reachable"

Prerequisites

• kubectl installed and on PATH
• One or more kubeconfig files (default: ~/.kube/config)
• Optional: kubectx/kubens, stern, helm

Investigation approach

The skill follows a read-first methodology: cluster health → namespace overview → resource describe → logs → events → hypothesis. It never suggests a mutating command before completing the read phase, and always requires explicit --context and --namespace flags to avoid silent mis-targeting.

Works well with

• argocd — GitOps sync and rollback after diagnosis
• dips-core:spector — check what version is deployed per environment
• dips-core:helm-qa — validate charts before re-deployment