FIX7 Authority-Seal — Codex Final Seal Review Packet

• Date: 2026-06-10 · Host: T1 · Codex consulted: NO · Production mutation: NO
• Lane: FIX7_AUTHORITY_SEAL_FULL_DRESS_REHEARSAL_AND_CODEX_READY_PACKET_MACRO_2026_06_10
• Status handed to Codex: FIX7_AUTHORITY_SEAL_FULL_DRESS_REHEARSAL_CODEX_READY
• This packet exists so Codex can author N7→N8→P7 inventing no field, order, tag, or encoding — every protocol element is fixed, executable, and self-red-teamed.

This is a REVIEW/ROUTING packet. It contains no real seal. T1 has authored nothing authoritative. The rehearsal digests are FIXTURE-only and labelled NOT-A-SEAL.

1. What T1 proved in this lane (full dress rehearsal)

Workstream	Evidence	Verdict
A. Fresh KB reconstruction	encoder re-fetched from governed KB, written fresh, shasum = 47200442f176b1c534f000c4079632f6388b17dd1763bdbac2cbb725a452b5bb, 19131 bytes; spec.json = f1c499270923c65c56783196b411ad6cacfe1188be0c531fabdd05a7795ef5eb, 4928 bytes	byte-exact, no local-only dependency
B. End-to-end rehearsal	authority_seal_rehearsal.py ran N7→N8→P7 (acyclic), deterministic 2nd pass identical	PASS
C. Byte-exact artifacts	per-node ordered records + preimage bytes + preimage_sha256 == digest; HASH_MANIFEST + packet_tree	PASS
D. Red-team (20 attacks)	authority_seal_redteam.py → 20/20 caught, 0 escaped	PASS
E. Spec/code/json/doc drift	authority_seal_drift_check.py → 22/22 agree, drift=0	PASS
F. Anti-hardcode/laundering	authority_seal_antihardcode.py → 9/9 PASS; broken-encoder flagged by drift oracle	PASS

Fresh-reconstruction packet tree: PACKET_TREE_SHA256=9f40519aa390497869850a12a82e73b2d3dd17ec53b5356b9c4fa1a243fb7314 (22 tracked files; excludes HASH_MANIFEST.txt, packet_tree.sha256, this cover doc).

2. Exact command Codex runs (no invention required)

cd <packet-root>                       # the dir holding authority_seal_encoder.py
shasum -a 256 authority_seal_encoder.py        # must equal 47200442…a452b5bb
python3 authority_seal_encoder.py --selftest   # must print 22/22 PASS, exit 0
python3 authority_seal_drift_check.py .        # must print 22/22 agree, drift=0, exit 0
python3 authority_seal_redteam.py rehearsal    # must print 20/20 caught, exit 0
python3 authority_seal_antihardcode.py         # must print 9/9 PASS, exit 0

Or simply bash rehearsal/commands.sh → ## RESULT rc=0.

Then, to author the real seal (acyclic order N7 → N8 → P7):

import authority_seal_encoder as E
n7 = E.encode_node("N7", [ ...A1/A2/A3/A5 + engineering N1..N6... ])  # -> envelope_manifest_sha256
n8 = E.encode_node("N8", [ ...N7 + Codex signer/timestamp/parent/report... ])  # -> detached_seal_sha256
p7 = E.seal_p7([ ...rev3 identity + Packet V3 tree + N7 + N8 + report/checkpoint/A1... ])  # -> authority_seal_pin_sha256

Field names, order, domain tags, and byte rules are fixed in authority-seal-encoder-spec.md §3/§4/§5 and authority-seal-encoder-spec.json.

3. Expected pass/fail conditions

Check	PASS condition	FAIL meaning
encoder sha256	== 47200442…a452b5bb	tampered/wrong encoder — STOP
--selftest	22/22 PASS, exit 0	contract regression — STOP
drift check	22/22 agree, drift=0, exit 0	spec/code/doc disagree — STOP
red-team	20/20 caught, exit 0	an attack escaped = contract defect — STOP
anti-hardcode	9/9 PASS, exit 0	digests may be laundered — STOP
N7/N8/P7 encode	returns 64-hex	a SEAL_* Reject names exactly the missing/extra/order/tag/cycle/constant/byte violation

Any Reject status is self-explaining; it tells Codex exactly which input to fix. No silent failure path exists.

4. Remaining TRUE authority inputs — what Codex MUST supply

Input	Node field	Actor	Notes
A1 approval_event_id	N7.approval_event_id, P7.approval_event_id	Codex	minted at the authorized approval event
A2 approver_identity	N7.approver_identity	owner + Codex	both parts supplied inside the seal event
A3 approval_event_timestamp	N7.approval_event_timestamp	Codex	stamped at seal time
A5 owner_blueprint_decision	N7.owner_blueprint_decision	owner	owner picks an option in owner-decision-packet.md §4 (OWN-1)
signer	N8.sealed_by	Codex	Codex signer identity
timestamp	N8.sealed_at	Codex	seal timestamp
parent	N8.parent_checkpoint	Codex	checkpoint document_id@revision
reports	N8.report_documents_digest	Codex	report_documents_digest([(doc,rev),…])
report/checkpoint ids	P7.codex_report_document, P7.codex_checkpoint_document	Codex	document_id@revision
engineering N1..N6	N7 sub-digests	engineering (already candidate)	from canonicalizer --produce / Packet V3

5. What Codex MUST NOT infer / invent

• No new field, field order, domain tag, or byte encoding — all fixed, proven byte-identical to canonicalizer rev3 (drift report).
• No prose-only P7 pin — rejected SEAL_PROSE_ONLY_PIN_REJECTED; P7 must be the computed digest (or the documented checkpoint-as-pin alternative, also byte-exact).
• No N7↔N8 cycle — N8 depends on N7, P7 depends on N7,N8; N7 never binds N8/P7. Any back-edge → SEAL_HASH_GRAPH_CYCLE.
• No treating a FIXTURE digest as a real seal — the rehearsal digests (6225f265…, b1f001b6…, 3599f663…) are FIXTURE outputs; the real seal will differ because real A1–A5/signer differ.
• No semantic guessing of signer/timestamp — the contract binds whatever Codex supplies and is tamper-evident; supplying the correct authority values is Codex's responsibility (semantic format validation is intentionally out of the encoder's scope — see red-team A7/A8).

6. Engineering boundary (unchanged)

• Packet V3 tree b95df0a5d2f41f80bea0cef8621c1f8bb0f6b49a40175116418494ed4141ca6d — unchanged, not redone.
• Canonicalizer rev3 49c386a9b9666c09786fc4f89bc79776b6046eaee6f4da6d8537d2c753b734d0 @ rev 3 / 38756 bytes — candidate, becomes authoritative only at the P7 seal.
• FIX7 implementation and all production/runtime gates remain BLOCKED (fix7-implementation-precondition-checklist.md).

7. Routing

Route the full closure packet (knowledge/dev/laws/tool-kiem-thu/packets/fix7-authority-closure-2026-06-10/) — encoder + spec md/json + n7/n8/p7 rev2 + owner-decision + precondition checklist + the four harnesses + rehearsal/ — to Codex for a new seal macro. Codex runs §2, supplies §4, observes §5, and authors N7→N8→P7. Nothing in the protocol is left for Codex to invent.