[asutermo]

Summary

• Expand rubric prefixes to DISCOVERY, AUDIT, ENVIRONMENT, and ANTI-CHEAT. Keep CREATE and all numeric and letter suffixes.
• Align canonical definitions, check IDs, scorer/report references, fixtures, and documentation. Preserve source order in CLI, demo, and retained reports.
• Preserve all 11 provenance criteria, 38 checks, 44 score anchors, and numeric source references. Keep the four anti-cheat criteria agent-guided.
• Version the renamed contract as rubric 4.1-draft, measurement output v3, and mutation-panel output v4. Scores remain 0 through 3 in the rubric; the bounded scorer still emits only 0 through 2 for the first three Discovery criteria.

Validation

• 256 affected tests passed on the final tree. Both independent reviews passed.
• Ruff lint and formatting, copyright headers, license checks, and Git whitespace checks passed.
• Canonical JSON matched the baseline exactly after identifier/version mapping. Source numbering, score anchors, and order were unchanged.
• Full local suite: 6 failed, 6860 passed, 126 skipped, and 76 subtests passed. All failures were in unchanged tests/test_verifier_cheat_evidence.py, which uses a 0.1-second subprocess timeout. Its 28 tests passed when rerun alone on both candidate and baseline main. The full suite is not reported as green.
• No live model or native-runtime acceptance was performed. Existing unassessed criteria remain unassessed.