ocp-triage-heuristicslisted

# OpenShift Triage Heuristics (JANUS-adapted) Experiential SRE decision knowledge for OpenShift/OCP investigations — the "what to check first, and why" that a seasoned operator carries. Distilled from the community `openshift-ops` plugin (marketplace `ecosystem-claude-plugins`, author Eran Cohen, Apache-2.0) and re-scoped to JANUS's read-only, sandboxed, human-approval discipline. ## How this differs from the source, and JANUS safety boundary `openshift-ops` targets a live operator who both diagnoses AND remediates. JANUS does not remediate autonomously. Every line below is tagged: - 🔍 **DIAGNOSTIC (autonomous-safe)** — read-only observation. A JANUS stage (lab-verify, live-tracer, doc-search, investigation-planner) may run/reason about it inside an approved disposable lab. Never against production. - ⚠️ **REMEDIATION (approval-gated)** — mutates cluster/node/operator state. JANUS NEVER executes these autonomously. They appear here only as *reference* so a report can recommend a fix for a human to apply, or so a hypothesis can be phrased ("stuck because a PDB blocks drain"). Treat as write-boundary: propose, never perform. ## Which JANUS stages consume this - **investigation-planner** — layered triage + failure-mode classes to decompose a symptom into tracks. - **lab-verify / live-tracer** — 🔍 diagnostic chains as read-only verification steps on the approved lab. - **doc-search / synthesize** — failure-mode vocabulary to correlate symptoms with CVE