FIX7 Real-N6 Candidate-Source Investigation (2026-06-11)

• Host: T1. Codex consulted: NO. Owner approval: NO. Production mutation: NO.
• Authority: provisional-non-authority; classification/evidence only.
• All sources live-read / fresh-reconstructed from governed KB this pass (not from prior prose).

A. Source authorization table

source	path	revision / hash	what it authorizes	what it does NOT authorize	implication for real N6
Canonicalizer SSOT (N2)	…/blueprint-2026-06-08/canonicalizer-fix7-canon-v1-ssot.md	rev — · md 49c386a9…b734d0 → py d9caa9fe…26f3e5	the algorithm that computes active_corpus_sha256	any value; any seal; any authority	gives a deterministic, governed way to compute N6
Recheck-9 Packet V2	…/packets/fix7-codex-recheck-9-2026-06-10/	tree pins via HASH_MANIFEST.txt rev2; produce evid rev1	byte-exact corpus + the on-record candidate d777e87c…	promotion; seal; official pin (Codex V3: engineering PASS, authority BLOCKED)	supplies the governed corpus + an expected value to cross-check
Authority-seal encoder	…/fix7-authority-closure-2026-06-10/authority_seal_encoder.py	rev2 · 13344f92…957144b8	the seal boundary (real/rehearsal provenance gates)	a real seal here (gated SEAL_REAL_N6_NOT_AVAILABLE)	proves N6 candidate ≠ seal; keeps N7/N8/P7 blocked
Rehearsal artifacts (S17)	…/fix7-authority-closure-2026-06-10/rehearsal/*	rev2/3	NOTHING — NOT-A-SEAL fixtures (6225f265… etc., corpus class REHEARSAL)	any real N6/N7/N8/P7	the thing a real candidate must not be (anti-laundering target)
N-number reconciliation	…/fix7-n-node-numbering-reconciliation-under-tkt-v02-2026-06-11.json	rev1 (PROPOSED)	engineering convention for labels	owner/Codex ratification; production approval	N6 label/semantics for this macro only
N6 special analysis	…/reports/fix7-n6-special-analysis-report-2026-06-11.md	rev1	clarification only	seal/promotion	decomposed the gap into engineering (T1) + authority halves

B. N6 candidate-source table

candidate	source path	source type	digest / value	provenance class	governed?	local-only?	can be real N6?	reason
active_corpus_sha256 (fresh recompute)	recheck-9 corpus (10 blueprint docs) via governed canonicalizer	computed-from-governed-bytes	d777e87c73d3b62d36789d9343f346102e98dbf301f2c93f7608470b876b258c	ENGINEERING_VERIFIED_CANDIDATE	YES (HASH_MANIFEST-pinned)	NO (reconstructed from KB)	YES	meets C1–C12; reproduced byte-exact this pass
active_corpus_sha256 (Packet V3 produce.log)	…/recheck-9/logs/produce.log / evidence/produce-expected-output.txt	report/log on record	d777e87c… (same)	ENGINEERING_VERIFIED_CANDIDATE (as cross-check oracle)	YES	NO	only as cross-check, not as the source	log prose is not the proof; the recompute is (C4)
rehearsal N6 fixture	…/rehearsal/n7-rehearsal-artifact.json & encoder FX["N6"]	NOT-A-SEAL fixture	6×64 (and N7 6225f265…)	REHEARSAL	n/a	n/a	NO	rehearsal class blocked (C2); laundering risk Codex flagged
membership_sha256	canonicalizer membership()	governed digest over doc ids	f2bda8…fe251	ENGINEERING (frozen pin)	YES	NO	NO (it is membership, not N6)	N6 is the corpus content proof, not the id-set pin
envelope_manifest (N7)	produce output	rehearsal-only marker	REHEARSAL_ONLY_NEEDS_SEALED_INPUTS	REHEARSAL/AUTHORITY	n/a	n/a	NO	N7 is an approval node, needs owner/Codex inputs
detached_seal (N8)	produce output	Codex-only marker	CODEX_ONLY_NOT_SELF_COMPUTABLE	CODEX_AUTHORED	n/a	n/a	NO	N8 authored by Codex at seal

Conclusion: exactly one source qualifies as a real N6 — the fresh recompute over the governed corpus, classed ENGINEERING_VERIFIED_CANDIDATE. Its value d777e87c… matches the on-record candidate (defence-in-depth), but the proof is the fresh byte-exact recompute, not the prior log.

C. Lane-boundary table

artifact / surface	owner lane	T1 may write?	T2 owns?	read-only?	object range	collision risk
FIX7 N-node / N6 / authority-seal semantics	T1	YES	no	no	this macro TKT-OBJ-241..253	none (>240 reserved ceiling)
Tool-Kiem-Thu v0.2 proof framework / packet	T2 (HOLD)	NO (must not mutate)	YES	read-only	—	n/a (not touched)
Recheck-9 Packet V2 / blueprint corpus	T1 (engineering baseline)	read-only (reconstruct)	no	read-only	existing	none (not mutated)
Authority-seal encoder + rehearsal	T1	read-only here	no	read-only	existing	none (not mutated)
Real-N6 provenance packet (NEW)	T1	YES	no	no	TKT-OBJ-241..253	none
Registry / 00-index	T1 governance	YES (addendum + pointer)	no	no	addendum	checked: max 225, reserved ≤240, new ≥241
PG / Directus / system_issues / production	—	NO	—	n/a	—	not touched

D. Investigation method (each load-bearing fact has first-hand evidence)

1. Fresh-reconstructed the recheck-9 tree + 10 corpus docs + governed encoder from KB (kb_fetch_reconstruct.py → RECONSTRUCTION: OK, 32 files + encoder, HASH_MANIFEST bidirectional).
2. Materialized the canonicalizer .py from the SSOT .md → d9caa9fe… (byte-identical to pinned).
3. Ran the governed --produce over the 10 docs → exit 0, byte-identical to pinned produce output, active_corpus_sha256 = d777e87c…, corpus_ok=True, membership_frozen_ok=True.
4. Confirmed the rehearsal fixtures are REHEARSAL-classed NOT-A-SEAL values, distinct from the computed candidate.
5. Confirmed (via the governed encoder) the seal path remains SEAL_REAL_N6_NOT_AVAILABLE in this lane.

No source was accepted on prose. No rehearsal value was promoted. No seal was authored.