k8s-log-investigation

What problem does it solve?

It helps you diagnose failing or unstable Kubernetes pods by quickly locating the most relevant error signals in very large log streams without dumping everything.

Core Features & Use Cases

• Token-efficient triage: searches for high-signal anchor patterns (errors, warnings, panics, timeouts, crash loop indicators) before retrieving broader context.
• Progressive widening: expands small surrounding context around each anchor, then gradually increases scope only as needed.
• Lifecycle correlation: checks previous container logs to connect failures with restart/crash behavior and improve root-cause confidence.
• Safe reporting: summarizes findings succinctly and avoids leaking secrets by redacting sensitive tokens; ends with an explicit result status line.

Use this when investigating pod crashes, CrashLoopBackOff, connection/timeouts, OOMKilled events, or when users ask to diagnose Kubernetes pod logs but the logs are too large to fetch in full.

Quick Start

Use the k8s-log-investigation skill to investigate the failing pod logs for the last 10 minutes in namespace <ns> and report the most likely root-cause category.