Debug Triage

A live system is broken and you do not yet know which layer is at fault. This skill localizes the failure - narrows an unknown-layer outage to one layer with evidence - and then hands off to the skill that owns that layer. It does not fix; it finds where to look.

It layers on top of the host harness's systematic-debugging skill: that skill is the root-cause method for one known component. Use this skill first, when the component is unknown, to decide which component systematic-debugging (or a domain skill) should then investigate. Triage localizes; the domain skill diagnoses; neither guesses a fix before the cause is found.

When to use

• A live service is down or degraded and the failing layer is not yet known ("it's broken", "502s started", "pods CrashLooping", "users report timeouts")
• An outage spans multiple layers (DNS, network, ingress, app, data, secrets) and you need to find which one before going deep
• Post-deploy or post-change breakage where the blast radius is unclear
• You are about to start guessing fixes - stop and localize first