Bind knowledge retrieval to sealed bytes and invalidate changed releases

Assistant: codex
Assistant-Model: gpt-6.1-sol
Assistant-Session: 01a10238-7ecc-78e2-903d-075f50e913cc

SCOPE.md

@@ -29,9 +29,9 @@ decisions without widening privilege, and keep root human-controlled.
 ## Current State
 
 - Canon and reference package: v0.8; Python 3.11+, PyYAML, pytest.
-- Knowledge package: v0.2.1 pilot; 14 cards, 24 owners, 69 pinned sources,
+- Knowledge package: v0.2.2 pilot; 14 cards, 24 owners, 69 pinned sources,
   read-only retrieval CLI, sealed exports and provider adapters.
-- Verification: 327 portable tests plus 9 optional real owner-contract tests.
+- Verification: 332 portable tests plus 9 optional real owner-contract tests.
 - Informed Decision unsigned adoption draft and native v0.2 orientation runbook;
   no live submission or execution bridge. See docs/decision-runbook-integration.md.
 - Thin pointers installed in four pilot repositories; fresh 0.2.0 pilot evidence
@@ -46,3 +46,12 @@ decisions without widening privilege, and keep root human-controlled.
 - Start with: INTENT.md
 - Agent instructions: AGENTS.md
 - Workplans: workplans/
+
+
+Owner reviews now provide scoped doctrine, vocabulary and PDP-boundary evidence
+for 0.2.1; their institutional/human adoption limits remain explicit. Whitehat
+reported two loaded-package consistency findings, addressed locally in 0.2.2
+with five regression cases. Whitehat review close-out was interrupted by a
+platform cybersecurity-risk block; no clean independent acceptance is claimed.
+Glas model evidence now leaves unreported observed models unknown. Current native
+continuity requirements remain in GLAS-WP-0012; distribution stays a draft.