chatbi-runbooklisted

# chatbi-runbook This Skill is the **procedural runbook** for the governed analysis flow routed by `/chatbi-analyze` (`.claude/commands/chatbi-analyze.md`). It encodes the deterministic order of clarification, source selection, degradation, quality check, independent review, gate enforcement, and footer assembly defined in `docs/chatbi-harness-domain-model.md` §10 and `docs/dev-cycle-3.md` §7. It is a **prompt / procedure artifact**, not executable code. Where a deterministic primitive exists in `.claude/lib/chatbi_harness/evidence.py`, this runbook names it; the actual offline exercise is Task 06 (`test_e2e.py`) and the real semantic-layer / adapter / reviewer runtime is a Cycle 5 exit gate. ## 0. Sources this runbook binds to - Request contract: `.claude/schemas/request.schema.json` (7 required fields). - Review verdict contract: `.claude/schemas/review.schema.json` (8 required fields, 11 coverage dimensions, `candidate_sha` binding). - Footer contract: `.claude/schemas/provenance.schema.json` (16 required fields). - Evidence primitives: `chatbi_harness.evidence.RunRecord`, `EvidenceEntry`, `compute_candidate_sha`, `validate_request`, `validate_review`, `validate_provenance` (fail-closed `GateError`). - Independent reviewer: `.claude/agents/adversarial-reviewer.md`. - Gates: `.claude/hooks/subagent_review_gate.py` (`SubagentStop`), `.claude/hooks/stop_gate.py` (`Stop`). ## 1. Step 1 - Clarify (Layer 1, REQ-001/002/003/004) **Goal:** before any query, confirm