What problem does it solve? Debugging Kubernetes workloads is slow when engineers jump straight to logs without understanding why pods entered a failing state. This Skill provides a structured investigation methodology that prioritizes cluster events over logs, dramatically reducing time to root cause for common pod and deployment failures. ## Core Features & Use Cases - Event-First Investigation Flow: Enforces a golden rule of checking pod events before logs, with a defined sequence from list_pods through describe_deployment. - Pattern-Based Diagnosis: Provides decision tables and checklists for CrashLoopBackOff, OOMKilled, ImagePullBackOff, Pending pods, probe failures, evictions, and stuck rollouts. - Error Classification: Distinguishes non-retryable errors (401, 403, 404, missing config) from retryable ones (429, 5xx, timeouts) to guide remediation. - Use Case: A pod enters CrashLoopBackOff after a deployment. Follow the flow to check events, identify an OOMKilled reason, compare memory usage against limits with get_pod_resources, and confirm the fix by raising the memory limit. ## Quick Start Ask the assistant to debug why the pods in the payments namespace are stuck in CrashLoopBackOff using the k8s-debug investigation flow.
