Claim validation hierarchy convergence workpackage

workpackage_claims/SCRUB-WP_REPOSITORY_CONVERGENCE_VALIDATION_HIERARCHY_CLARIFICATION.md

@@ -0,0 +1,52 @@
+# Workpackage claim — SCRUB-WP_REPOSITORY_CONVERGENCE_VALIDATION_HIERARCHY_CLARIFICATION
+
+Status: implementation in progress  
+Role: `implementation_operations`  
+Issue: #117  
+Branch: `wp/repository-convergence-validation-hierarchy`  
+Starting main: `7e4f5491fa6616f9f1b08649a4ed9dfd80de0d84`  
+Started: 2026-09-05 Europe/Amsterdam
+
+## Objective
+
+Remove ambiguity about validation authority without creating another framework or changing any product/runtime behavior.
+
+## Source-derived finding
+
+Current evidence has three different roles that historical documents sometimes blur:
+
+1. `.github/workflows/tests.yml` runs the complete committed `tests/` suite on pull requests and `main`; this is the current exact-SHA regression gate.
+2. Capability/domain suites inside `tests/` (Phase-6 E2E, Scrub Key, document fidelity/hygiene, Zorg, Premium/AppTest, recognizer contracts) provide specific product-regression evidence as part of that full suite.
+3. Recall/precision/report helpers are diagnostic evidence, not merge/production gates: `recall_benchmark_report.py` explicitly sets `diagnostic_only`, `production_gate=False`, `thresholds_enforced=False`; WP22 scores supplied predictions and does not invoke recognizers; WP23/WP24 are report-only and non-gating.
+
+## Scope
+
+Authorized:
+
+- record one binding validation-authority decision in `DECISION_LOG.md`;
+- align the validation strategy wording in `ROADMAP.md`;
+- advance `WORKPACKAGES.md` from the merged WP-CONVERGENCE-02 to this package;
+- align the temporary convergence debt ledger;
+- update `CHANGELOG.md`;
+- add one narrow contract test that validates workflow/source metadata and the binding decision;
+- write the mandatory handover.
+
+Not authorized:
+
+- new Evidence Framework or validation subsystem;
+- changes to corpus data, recognizers, thresholds or product/runtime code;
+- making diagnostic recall/precision scores release-blocking;
+- production-safety or perfect-recall claims;
+- deleting historical validation helpers merely because they are non-authoritative;
+- issue-state cleanup unrelated to this validation-authority cluster.
+
+## Acceptance
+
+1. One release regression gate is unambiguous: exact-SHA full `Tests` workflow.
+2. Capability/domain tests remain first-class regression evidence but not competing merge authority.
+3. Recognizer-backed recall report is explicitly supplemental diagnostic evidence, consistent with its source metadata.
+4. WP22/WP23/WP24 scorecard/residual-risk chain is classified supplemental/historical diagnostic and not release authority.
+5. No new threshold or production-safety claim is introduced.
+6. Full exact-head `python -m pytest -q tests` is green.
+7. Fresh independent `governance_release_assurance` returns PASS before merge.
+8. Post-merge exact-main Tests and GitHub→Hugging Face sync are verified where applicable.