# Exact unsent Grok prompt — EM-RSK-02

Independent enterprise metamodel review. Do not browse, invent identifiers, give legal advice or claim standards conformance.

EM-RSK-02 covers Audit, AuditProcedure, AuditFinding, AssuranceOpinion and CorrectiveAction using WM-ACT-033 Review / Inspection / Audit, WM-ECO-035 Audit / Assurance Engagement and WM-KNW-014 Issue / Problem, with assessment, claim, task, rule and risk/control boundaries.

Assess this proposal: WM-ECO-035 masters engagement/mandate/opinion while WM-ACT-033 masters execution/finding, but their dual-root overlap must be adjudicated before publication. A reusable ProcedureDefinition is an identifier-unassigned candidate; a performed test is execution-local. AuditFinding reuses WM-ACT-033 and aligns to WM-KNW-007 without duplicate identity. AssuranceOpinion is an engagement-scoped statement, not a new model. CorrectiveAction profiles WM-ACT-006; management response is separate. Keep observation, evidence, finding, issue, risk and task distinct. Pin criteria and sampling frame/method/sample; scope conclusions and exclusions; require independent retest for closure. Task completion never closes a finding, and opinion never proves general compliance.

Test two processes with one excluded site, explicit frame gaps and a finding retested after corrective action. Reject extrapolating one-system evidence to group-wide conformity.

Return <=900 words with: Verdict; strongest evidence; strongest counterexample; identity/mastership; engagement/execution/procedure; sampling/evidence; finding/issue/risk; opinion/scope; corrective action/closure; independence; scenario; at least 10 invariants; minimum model set; blockers. Explicitly decide which candidate needs independent identity and how to resolve the dual-root conflict.