FIX7 Digest / Seal Reproduction Report

• Date: 2026-06-10 · Object ID: TKT-OBJ-055 · provisional-non-authority, evidence-only · Codex: NOT consulted
• Encoding source: doc 00 §"Canonical hash encoding / Per-aggregate specs / extractor" (NON_AUTHORITY_EXPLANATION of the SSOT). All 7 non-membership values are SEAL_AT_CODEX_RECHECK_8 by design.
• Evidence (hash-bound + rerunnable): …/packets/fix7-codex-recheck-9-2026-06-10/ README §5, RERUN.sh step 4.

1. Each digest: input, computation, expected, actual, class

#	digest	node	tag / record encoding	expected	actual (candidate)	PASS/REHEARSAL/BLOCKED
1	membership	N2	FIX7_ACTIVE_AUTHORITY_MEMBERSHIP_V1 · record=document_id · sort asc	f2bda8…fe251 (frozen)	f2bda8…fe251	PASS (frozen reproduced; shasum==hashlib)
2	normalized_active_content ×10	N1	FIX7_DOC_NORMALIZED_CONTENT_V1\t<id>\n + normalized active bytes	deterministic	10 hashes (00=b1665a…45602B, 12=39a6c0…47818B, 06=1144b7…)	REAL CANDIDATE (seals at Codex)
3	active_corpus	N6	FIX7_ACTIVE_AUTHORITY_CORPUS_V1 · record=id\tstatus\tsection\tkb_revision\tN1 · sort asc id	deterministic	d777e8…	REHEARSAL — kb_revision fields seal at Codex (placeholder SEAL_AT_CODEX_RECHECK_8)
4	marker_fence_registry	N3	FIX7_MARKER_FENCE_REGISTRY_V1 · record=id\tkind\tliteral · sort asc	deterministic	bb9ca0…	REAL CANDIDATE
5	superseded_boundary	N4	FIX7_SUPERSEDED_BOUNDARY_V1 · record=<id>#S<k>\tL<b>-L<e> · sort asc	deterministic	9b111c…	REAL CANDIDATE
6	guard_set	N5	= N1(doc 06)	=doc06 N1	1144b7… (== doc06 per-doc hash)	REAL CANDIDATE (wiring confirmed)
7	canonicalizer_sha256	—	sha over SSOT .md MCP bytes (CRLF/CR→LF)	hash of SSOT bytes	8f80f9… (local copy)	CANDIDATE (authoritative=KB bytes at sealed rev)
8	envelope_manifest	N7	FIX7_ACTIVE_AUTHORITY_ENVELOPE_MANIFEST_V1 · 24-field roster + per-doc tuples + superseded list + anchor flag	—	REHEARSAL_ONLY	BLOCKED — binds sealed sub-digests + approval-event fields (approved_status/epoch/by_role/at_utc/parent_recheck_checkpoint) only Codex/owner set
9	detached_seal	N8	FIX7_CODEX_DETACHED_SEAL_V1 · seal fields excl. self+signature	—	CODEX_ONLY	BLOCKED — Codex authors sealed_by/at, signature, parent_checkpoint_id, report_documents[]

2. Extractor reproduction proof (per-doc N1 inputs)

The deterministic extractor was applied to all 10 byte-exact docs and ran clean (zero fail-closed errors): each doc has exactly one DOC_STATUS: ACTIVE_AUTHORITY; doc 00 (self-host) dropped its ENVELOPE:EXCLUDE region + 1 superseded fence (61577→45602 active bytes); doc 12 dropped 2 superseded fences (65178→47818). CRLF/CR→LF normalization, flat fence pairing, and inclusive removal are unit-pinned in the 36-check selftest.

3. DAG mapping (acyclic, recheck-6/7 accepted)

N1(per-doc)→N6(corpus); N2,N3,N4,N5,N6,N1→N7(manifest, binds NO checkpoint hash, NOT N8); N2,N5,N6,N7→N8(seal, not self, not checkpoint hash); N9 diagnostic sink. Acyclicity asserted in selftest (and a self-edge is detected as a cycle).

4. Why "all 8 produced" is NOT achievable in this macro (precise blocker)

N7 and N8 are seal-event digests by construction: their inputs include approval metadata and Codex authorship that do not exist until the seal. Computing them now would mean hashing placeholder/fabricated approval fields — which the blueprint explicitly forbids ("self-fabricated authority"). Therefore the honest maximum is 6/8 computable candidates + 2/8 produced-at-Codex-seal. This is the Codex seal handoff (P7), not an engineering deficiency.

Verdict

DIGEST_SEAL_REPRODUCTION_PARTIAL_BY_DESIGN — 6/8 aggregate digests + 10 per-doc hashes reproduced deterministically with real inputs; N7/N8 precisely blocked on Codex/owner seal inputs; no value fabricated; reproducible via RERUN.sh.