KB-4B8A

T1 FIX7 Blueprint Patch After Codex Critical Fail - Readme First

5 min read Revision 1
fix7codex-patchreadme2026-06-08

00 - Readme First: T1 FIX7 Blueprint Patch After Codex Critical Fail

Date: 2026-06-08 Author: T1 (production Agent for Agent Data) Macro: PROGRAM_PATCH_T1_FIX7_REFACTOR_BLUEPRINT_AFTER_CODEX_CRITICAL_FAIL Mode: READ-ONLY production. Blueprint KB-doc direct-revision allowed. No production mutation.

What this is

Codex independently critical-reviewed the FIX7 Existing-System Refactor Execution Blueprint and FAILED it: FIX7_REFACTOR_BLUEPRINT_CODEX_CRITICAL_REVIEW_FAIL_HARDCODE_OR_PG_NATIVE_GAP, with 7 blockers and an A-K verdict matrix. This package is T1's repair: it patches every Codex blocker directly in the blueprint documents, then self-reviews against the same Codex checks.

This is NOT implementation, live migration, Stage 2.6B, a permit, REAL_RUN, QT001 apply, manifest activation, owner/ACL cutover, or any production mutation. The only writes are KB blueprint-doc revisions, this patch report, and the patch checkpoint.

Final status

FIX7_REFACTOR_BLUEPRINT_T1_PATCHED_AFTER_CODEX_FAIL_READY_FOR_CODEX_RECHECK

The 7 Codex blockers and where each was fixed

# Codex blocker primary fix blueprint docs patched
1 legacy target set = disguised hardcode (owner+name pattern) sealed legacy-disposition set in authority_scope_manifest #20 (typed identity + source/priv hash + exact-set both-EXCEPT vs catalog); name/owner scan = diagnostic candidate only; +G-LEGACY-TARGET-SEALED; G-PGNATIVE extended 01, 02, 03, 04, 06, 07, 08
2 G-NOLEGACY phase-deadlock split G-NOLEGACY into PRE (gates PKG-F) + POST (verifies PKG-F) 04, 06, 07
3 stub-scope contradiction ("stub all" vs "apply/writer only") 5 dispositions; stub only STUB_FAIL_CLOSED; symmetric, body-bounded rollback 02, 04, 05, 06, 07, 08
4 rollback may reopen PUBLIC EXECUTE with new plane present atomic deactivation-first rollback + G-NOMIXED-AUTHORITY 04, 05, 06, 07
5 ACL snapshot incomplete +column ACL + effective role membership + sequence/default/PUBLIC/Directus/cp grants + snapshot hash; both-direction effective-privilege verify 05, 06, 07
6 writer-gateway identity/owner ambiguous at S15/S16 pinned #26 identity + phase-explicit owner table; +G-WRITER-GATEWAY-IDENTITY 02, 04, 06, 07, 08
7 "permit" ambiguous vs operator authorization operator_authorization (machine-checkable) vs blocked qt001_backfill_permit vs REAL_RUN_authority; +G-NO-QT001-PERMIT-DURING-FIX7; law §4G citation corrected 03, 04, 07, 08, 12
8 A-K verdict alignment A-K mapping table in this report + blueprint doc 12 12 + report doc 01

Live grounding (read-only, DB directus, 2026-06-08, query_pg)

The patch is grounded on fresh live evidence, not the prior PASS claims. The decisive BLOCKER-1 finding: the legacy "set" changes with the name literal -

  • views: qt001% prefix → 0, v_qt001% prefix → 183, %qt001% substring → 196;
  • routines: %qt001% substring → 45 functions + 1 procedure (= 46 routines), all owner directus, all proacl=NULL (PUBLIC EXECUTE), prosecdef=false; 0 qt001 routines outside public;
  • tables: 20; qt001_cp schema + 3 roles still absent.

A set that swings 0/183/196 by literal cannot be operational authority. The blueprint now binds the target as a sealed, hash-bound, exact-set #20 manifest, not a pattern.

Document map

doc content
00 this readme
01 Codex failure matrix (A-K) + blocker→fix→docs→self-check
02 BLOCKER 1 - legacy target authority fix (sealed #20 disposition set)
03 BLOCKER 2 - G-NOLEGACY phase fix (PRE/POST)
04 BLOCKER 3 - stub-scope disposition fix (5 dispositions)
05 BLOCKER 4 - rollback no-mixed-authority fix (atomic deactivation-first)
06 BLOCKER 5 - ACL snapshot completeness fix (column ACL + effective privilege)
07 BLOCKER 6 - writer-gateway identity fix (pinned, phase-explicit)
08 BLOCKER 7 - permit terminology fix (operator_authorization vs blocked permit)
09 self-review: hardcode / disguised-hardcode / PG-first-native-driven
10 self-review: cross-layer boundary
11 direct blueprint revisions applied (every patch, doc, revision)
12 final verdict

Boundary status (unchanged by this patch)

Official FIX7 design: approved. Blueprint: patched, pending Codex recheck. Implementation, Stage 2.6B, qt001_backfill_permit, REAL_RUN, QT001 apply, manifest activation, repoint, owner/ACL cutover: ALL BLOCKED. Next after PASS is Codex recheck only - not implementation approval.

Back to Knowledge Hub knowledge/dev/reports/architecture/t1-fix7-blueprint-patch-after-codex-critical-fail-2026-06-08/00-readme-first.md