KB-3EB5

Codex FIX7 N-number / N6 / Authority Seal Review — REJECT

10 min read Revision 1
fix7codexn-numbern6authority-sealreject2026-06-11

Codex FIX7 N-number / N6 / Authority-Seal Review

  • Host: CODEX
  • Date: 2026-06-11 Asia/Ho_Chi_Minh
  • Final status: CODEX_FIX7_N_NUMBER_OR_N6_REJECTED
  • Production mutation: NO
  • N7/N8/P7 authored: NO / NO / NO

Executive decision

The N6 digest d777e87c73d3b62d36789d9343f346102e98dbf301f2c93f7608470b876b258c is reproducible from the governed corpus, but it is not ratified as an ENGINEERING_VERIFIED_CANDIDATE in this review because Codex-owned probes found two fail-open duplicate classes that emit a certificate.

The proposed N-number table is also not ratified for binding use because a load-bearing authority document still identifies membership_sha256 as N1, while the proposed table defines N1 as the per-document normalized-content digest and makes membership unnumbered. Ratifying the proposal without reconciling the active authority surfaces would preserve conflicting binding meanings.

T2 and T1-alignment review packets additionally fail fresh governed-KB reconstruction because manifest-listed load-bearing files are absent from their packet paths.

Answers required by the macro

Question Verdict
N-number canonical table safe to ratify for binding use? NO — active authority-envelope conflict plus incomplete alignment packet.
N6 digest real/non-rehearsal and reproducible? Digest reproducible, but candidate NOT ratified because verifier fails open on duplicate declarations/manifest entries.
T1/T2 proof packets reconstruct and fail closed? NO — T2 and alignment packets do not reconstruct; real-N6 packet reconstructs but fails Codex-owned duplicate probes.
Evidence laundering found? No rehearsal/local/prose laundering was accepted in tested paths, but duplicate ambiguity can be certified as governed evidence.
All authority inputs present? NO. A1/A2/A3/A5, owner decision, N8 signer/report/checkpoint inputs, and P7 report/checkpoint inputs are absent.
N7/N8/P7 authorable now? NO. Not authored.

File reality table

Path/surface Exists? Hash/tree Role Authority level Load-bearing Verified by command
Real-N6 packet YES tree 43b4914acd52a1c410ac1fb4cd1f09d676f3ab53c4af99204e65c15d752eea23 N6 evidence provisional non-authority YES YES
Alignment packet PARTIAL tree text matches manifest hash 96d00b9e..., but 2 manifest files absent N-number/encoder alignment non-authority YES YES, FAIL
T2 v0.2 review packet PARTIAL tree text matches manifest hash b221f224..., but 6 manifest files absent proof convention non-authority/not promoted YES YES, FAIL
Governed reconstructed encoder YES 13344f92cafcaf0d07dcb21700bdb642f38b89351702e08080eacb0e957144b8 seal contract boundary provisional contract YES YES
Canonicalizer algorithm YES d9caa9fe9f46854c38c996747d50d2e73bc5074705730e869fd6b1f8cc26f3e5 N6 computation engineering YES YES
Canonicalizer SSOT YES 49c386a9b9666c09786fc4f89bc79776b6046eaee6f4da6d8537d2c753b734d0 candidate pin engineering candidate YES YES
N6 certificate YES file acf8d259...; binding 055828db... candidate certificate NOT_A_SEAL YES YES, but verifier defect invalidates ratification

N-number table

Node Proposed canonical meaning Source Conflict? Load-bearing Ratifiable? Reason
N1 normalized_active_content_sha256[d] proposed reconciliation YES YES NO Authority N7 envelope still labels membership_sha256 (N1).
N2 canonicalizer_sha256 proposed reconciliation historical conflict YES NO Binding packet incomplete; all active surfaces are not reconciled.
N3 marker_fence_registry_sha256 model NO found YES held Cannot ratify table partially as binding.
N4 superseded_boundary_sha256 model NO found YES held Same.
N5 guard_set_sha256 model cosmetic ambiguity recorded YES held Same.
N6 active_corpus_sha256 model/recompute verifier defect YES NO Duplicate inputs can receive a certificate.
N7 envelope_manifest_sha256 seal contract NO semantic conflict found YES NO Upstream not ratified; authority inputs absent.
N8 detached_seal_sha256 seal contract NO semantic conflict found YES NO N7 absent and Codex inputs absent.
N9 diagnostic sink proposal/model historical naming conflict NO held Not authority-bearing, but full table is not ratified.
membership unnumbered active_corpus_membership_sha256 proposal YES YES NO Authority envelope still calls it N1.
P7 authority_seal_pin_sha256 authority layer NO semantic conflict found YES NO N7/N8 absent.

N6 table

Digest Source corpus Provenance Governed Reconstructable Rehearsal Ratifiable Reason
d777e87c...b258c 10 governed blueprint documents claimed ENGINEERING_VERIFIED_CANDIDATE YES YES NO NO Verifier accepts duplicate declared members and duplicate manifest entries, then emits a certificate.
6 x 64 fixture rehearsal fixture REHEARSAL fixture only YES YES NO Correctly rejected from real path.
f2bda8...fe251 membership set engineering membership pin YES YES NO Not N6 Membership digest is not active corpus digest.

Authority input table

Input Required for Present Source/class Can Codex author now? Missing action
Ratified N-number table N7/N8/P7 binding NO proposed only NO Reconcile active surfaces and rerun review.
Ratified real N6 candidate N7/N8 NO reproducible but verifier-defective NO Close duplicate fail-open and rerun.
A1 approval_event_id N7/P7 NO AUTHORITY_INPUT NO Authorized seal event required.
A2 approver_identity N7 NO owner + Codex NO Owner/Codex decision required.
A3 approval timestamp N7 NO AUTHORITY_INPUT NO Mint at authorized seal event.
A5 owner blueprint decision N7 NO owner NO Owner must select/authorize.
N8 signer/timestamp/checkpoint/report digest N8 NO CODEX_AUTHORED NO Only after valid N7.
P7 report/checkpoint/N7/N8/A1 P7 NO CODEX_AUTHORED/authority NO Only after valid N7/N8.

Independent fail-open findings

Codex supplied invalid inputs beyond T1 tests:

  1. declared_members = CORPUS_MEMBERS + [CORPUS_MEMBERS[0]] was accepted by verify_real_n6() and emitted a certificate-like dictionary. Expected N6_CORPUS_DUPLICATE.
  2. A HASH_MANIFEST.txt with a duplicate docs/... entry was accepted and emitted a certificate-like dictionary. Expected a duplicate-manifest rejection.

These violate C8 and the macro rule that any accepted invalid input requires rejection. T1's case-variant directory-listing probe does not cover duplicate declared membership or duplicate manifest records.

Reconstruction verdict

  • Real-N6 packet: RERUN_RESULT: PASS; fresh governed KB reconstruction succeeded; canonical gates passed.
  • Alignment packet: RERUN_RESULT: FAIL; governed packet paths lack authority_seal_encoder.py and codex_probes.py despite manifest entries.
  • T2 review packet: OVERALL_RESULT: FAIL; governed packet paths lack content_bind_oracle.py, content_bind_verify.py, content_bind_regression.py, content_bind_rerun.sh, content_bind_oracle.json, and real-sut/fix7_canon_v1_ssot.py.

Authority firewall and authoring verdict

The tested rehearsal/local-only/OFFICIAL_PIN-without-authority paths rejected correctly. Missing N7/N8/P7 provenance also rejected correctly. Nevertheless, the authority firewall is not sufficient because duplicate ambiguity is certified at the N6 evidence boundary.

No N7, N8, P7, approval, official pin, Tool-Kiem-Thu promotion, implementation, or production action was authored.

Blockers

ID Failed/missing input Evidence Actor Next action Blocks
CODEX-N6-DUP-DECLARED Duplicate declared corpus member accepted Codex-owned probe emitted certificate T1 Reject duplicate declared members before set conversion; add regression probe. N6, N7, N8, P7, implementation, production
CODEX-N6-DUP-MANIFEST Duplicate manifest record accepted Codex-owned probe emitted certificate T1 Reject duplicate manifest paths and malformed records; add regression probe. N6, N7, N8, P7, implementation, production
CODEX-NNODE-ACTIVE-CONFLICT membership_sha256 (N1) remains in authority N7 envelope governed n7-approval-event-input-envelope.md rev3 vs proposed table T1/owner/Codex Reconcile every load-bearing authority surface, then request binding ratification. N-number ratification, N7 onward
CODEX-ALIGNMENT-PACKET-INCOMPLETE Two manifest-listed files absent fresh fetch + shasum -c / RERUN.sh exit 1 T1 Publish actual files or a governed self-contained reconstruction that passes bidirectionally. proof packet, ratification
CODEX-T2-PACKET-INCOMPLETE Six manifest-listed files absent fresh fetch + commands exit 1 T2 Publish/reconstruct actual files at governed paths; preserve non-authority status. proof-convention verification
AUTHORITY-INPUTS-MISSING A1/A2/A3/A5, N8/P7 Codex inputs and owner decision absent governed authority request docs Owner/Codex Supply only in a later authorized seal macro after engineering blockers close. N7/N8/P7

T1 should patch only the real-N6 duplicate validation and reconcile all binding N-number references. T2 should make its review packet reconstructable without promotion. Then route a new independent Codex review macro. Production and implementation remain blocked.

Back to Knowledge Hub knowledge/dev/reports/architecture/codex-fix7-n-number-n6-authority-seal-review-2026-06-11/00-readme-first.md