What problem does it solve? When pods enter CrashLoopBackOff, ImagePullBackOff, or OOMKilled states, on-call engineers must manually correlate pod status, logs, events, and deployment history to find the root cause. This Skill automates that investigation workflow and produces a structured diagnosis report with severity classification and remediation recommendations. ## Core Features & Use Cases - Systematic Cluster State Collection: Gathers pod status, container logs (including previous crashed instances), warning events, and deployment rollout history via kubectl. - Decision-Tree Diagnosis: Classifies failures by reason code (CrashLoopBackOff, ImagePullBackOff, OOMKilled, high restart counts, stuck rollouts) and maps evidence to root causes. - Safe Remediation Guidance: Recommends specific kubectl actions such as rollout undo or memory limit increases, with strict escalation rules requiring human approval for any mutation. - Use Case: An on-call alert fires for a pod in CrashLoopBackOff. Provide the namespace, and the Skill collects evidence, identifies an OOM kill from exit code 137 in the logs, and recommends increasing memory limits pending approval. ## Quick Start Investigate the unhealthy pods in the payments namespace and tell me the root cause with recommended fixes.
