ops-debugginglisted

# Ops & Deployment Debugging ## What it does Generic debugging methodology (repro-minimise, hypothesis discipline, instrumentation) lives in `/ms.fix`'s Hard-Bug Discipline. This skill owns a narrower, evidence-based slice: failures that only manifest at an environment boundary — a deploy, a tunnel, a proxy, a container, a rotated secret, a platform-native API — where the fastest path to a fix is recognizing the failure *class*, not re-deriving it from first principles every time. Every entry below traces to a real diagnosed incident, stated at the failure-class level with project specifics stripped — this is not a generic textbook list. If a failure class isn't below, fall back to `/ms.fix`'s Hard-Bug Discipline. ## When to use - A symptom only reproduces after deploy, never locally (or vice versa). - "It worked yesterday" / "no code changed" with a new failure. - A network boundary is involved: SSH, tunnel, reverse proxy, TLS, DNS. - A container won't start, is stuck `unhealthy`, or a bind-mount looks wrong. - Auth/login breaks specifically around a secret rotation or redeploy. - A real-time/streaming API integration (WebSocket, SSE) rejects a connection or silently stops producing output. ## Method (apply before reaching for a specific entry below) - **Read the actual process-level error, not the surface signal.** An orchestrator-level warning, a familiar-looking symptom, or an assumption about which host a check ran on is very often wrong or incomplete. The l