What problem does it solve? When something breaks — a failing service, a crashing pod, a misbehaving script — ad-hoc guessing wastes time and introduces new problems. This Skill enforces a disciplined debugging methodology: gather evidence first, form a ranked hypothesis, test one change at a time, validate the fix, and record findings so the next investigation starts smarter. ## Core Features & Use Cases - Structured 7-step methodology: Understand the system, gather evidence (logs, processes, resources, network, config, recent changes), form and rank hypotheses, test with minimal probes, apply the smallest fix, validate, and record findings. - Parallel investigation: Automatically dispatches investigator sub-agents when multiple independent targets or leads exist (multiple pods, services, or environments), keeping serial work only for dependent or mutating steps. - Cross-platform command coverage: Ready-to-use diagnostic commands for journald/systemd, Docker, Kubernetes, network tools (curl, nc, dig), and resource inspection (top, df, lsof). - Use Case: A Kubernetes deployment starts crash-looping after a config change. The Skill walks you through pulling pod logs and previous-container logs, checking recent git history, isolating the bad config value, applying a minimal fix, verifying rollout status, and logging the root cause to the project knowledge base. ## Quick Start Ask the agent to debug why the API service is returning 500 errors and follow the evidence-first investigation workflow.
