DECISION_DOSSIER.md

decisions/DECISION_DOSSIER.md

Batch-1 Decision Dossier

Every open ledger decision and my questions for you, worked one by one: options,

my best recommendation, how it fits the CVD, how it builds canon, why this option,

the surface/mid/deep (L1/L2/L3) player analysis, a concrete test, and a

fresh-context adversarial review of each recommendation. Produced by a 20-decision

ultracode workflow (deep-dive research -> recommendation -> independent adversary),

then assembled deterministically. Grounded/mechanical items already applied this

session are marked APPLIED; canon-invention / hard-line / naming / numerical calls

stay AWAIT — yours to rule.

Summary table

---

1. Approve the 75-row boss roster with three narrow row-level amendments

• Key: 0.E-boss-roster Type: creative Disposition: await_josh_confirm

Options

• Approve as-is — Ratify all 75 rows exactly as staged (Candidate B synthesized) and populate T0_Boss_Encounter after the personal-name/D-9/region flags are ruled, with no further changes.
• Canon support: The roster's core design verifies against direct canon reading: blended-never-single-mode with the Ch 17 three-phase template as structure [SRC: CVD L214-218 Pillar 8; SRC: COMBAT L207/L372]; hard-line chapters carry a real playable combat layer ("combat absence is creatively fatal") [SRC: CVD L181-183 Pillar 3; SRC: COMBAT L371,L495]; environmental 4-step protocol bosses match the documented instances [SRC: COMBAT L336-356]; graduated-consequence failure states are canon and drive the Hades loop [SRC: COMBAT L343-347]; the five Tier-A anchors Ch 8/23/38/55/65 are confirmed and the deleted-"anchor #5" correction is right [SRC: COMBAT L211-215,L495; SRC: T4 L389 Ch 58 = first direct engagement, Cassius absent, no fifth-anchor sentence]. AGAINST: the roster's OWN executable anti-staleness rule (no two consecutive chapters share dominant combat type; Ch 40-41 the sole exception) is violated by Ch 04-05 (strike/strike), Ch 43-44 (infiltrate/infiltrate), and a Ch 37-38 attune/attune label collision [SRC: CVD L232 Pillar 10 executable test; SRC: CVD L664/L879 §12.2/§16.2; SRC: T4 L229 sole exception].
• Approve-with-amendments (RECOMMENDED) — Ratify the roster as the winning design, and apply three narrow row-level fixes before population: (a) break the three unsanctioned consecutive same-dominant-type pairs by promoting an already-present secondary verb to dominant on one side of each; (b) elevate the Ch 57 fauna item from a design-neutral flag to an explicit Josh ruling, recommending the more-protective T4 reading; (c) keep the personal-name TBDs, D-9, Ch 43 region, and revisit-order as the standing Josh-exclusive flags. No structural rework.
• Canon support: Each amendment is a grounded correction, not an invention. Ch 04 already carries perceive ("find the coercer behind the screen") and Ch 05 already carries decode ("1025 CE Chola-Srivijaya evidence trail") — canon supports a non-melee dominant on either [SRC: T4 L271 Ch 04 certification-control; SRC: T4 L273 Ch 05 institutional-destruction/decode]. Ch 43 carries expose and a part-break win alongside infiltrate; Ch 44 stays infiltrate — promoting Ch 43's expose breaks the stealth/stealth pair with zero canon cost. Ch 37/38 already annotate distinct dominant MODES (traversal-endurance vs seven-vril-seat protocol) so a label fix satisfies §12.2 "mode emphasis changes per chapter" [SRC: CVD L666]. The Ch 57 item is a real T4-vs-COMBAT contradiction: T4 L387 says "no fauna combat"; COMBAT L371/L495 lists "person-to-animal with documented dangerous fauna" — under CVD > T1-domain authority and §17.1 maximum protection, the more-protective no-fauna reading should win [SRC: CVD L933/L181 §17.1 & Pillar 3; SRC: T4 L387; SRC: COMBAT L371].
• Rework — Send the roster back for re-authoring (new candidate pass or major structural revision).
• Canon support: Not warranted by the evidence. The roster passes every load-bearing canon check I verified directly (Pillars 3/8/10/11, §12.2/§12.3, COMBAT §7/§9, T4 anchor set and Ch 58 correction). The only defects found are three row-level label collisions and one flag-elevation — all fixable in place. Rework would discard a verified-winning design for cosmetic fixes [SRC: CVD L214-247; SRC: COMBAT L202-374].

Recommendation — Approve-with-amendments (RECOMMENDED)

How it fits the CVD. The roster is the CVD's boss doctrine instantiated, and it holds up under direct reading rather than summary trust. Pillar 8 (blended, never single-mode) is satisfied per row: every row carries primary+secondary verbs, a three-axis Patterns/Puzzles/Preparation split that is never 100/0/0, and a character-engagement field — matching the executable test [SRC: CVD L214-218] and the Ch 17 Entu three-phase template named as canonical structure [SRC: COMBAT L207,L372]. Pillar 3 hard-line touchability is honored the way CVD demands: Ch 29/38/42/47/48/50/57/69 route combat through Velheim intrusion, constructs, or the astral/dream layer, never against cultural representatives, with a real playable layer at every protected chapter (\"Hard Line chapter without combat layer at all = pillar violation; combat absence is creatively fatal\") [SRC: CVD L181-183]. Ch 69's materialized argument is built as a genuine playable ACT-class boss (declare-verbs + perceive/dodge positioning, weapons canonically unresponsive), which is exactly the CVD's stated gameplay product for the Architect arc [SRC: CVD L175-177]. Pillars 10/11 mode variety is treated as an executable roster-wide check, and §12.3's environmental-boss mechanic (perception/vril-management/restraint, victory not measured in damage) is faithfully rendered at the 9+2 documented and staircase instances [SRC: CVD L668-676; SRC: COMBAT L336-356]. The one place the roster falls slightly short of its own CVD-derived rule is the three consecutive same-dominant-type pairs — which the amendments close.

How it builds canon. Approving this instantiates the empty flagship T0_Boss_Encounter registry (0 rows today) with 75 canon-anchored rows, which is precisely the foundation-population Known-issue #4 says the spine cascade needs before/early. It stands up the mandatory Pillar 8 boss fields (primary/secondary verb, axis split, phases, escalation vector, win condition, fairness, reward, mastery tier) so the Pillar 8/10/11 audits become runnable per chapter instead of aspirational [SRC: CVD L218-219 mandatory fields; SRC: AAS/registry Boss_Encounter empty]. It also locks the verb-curriculum spine (new verbs taught in valleys, tested at the next peak; the purge lineage 26->58->70->72->74; the declare-lineage Ch 17->65->69) that the /spine-chapter cascade consumes as its combat-encounter load manifest. Because every row is anchored and only the game-design shaping is authored, it advances canon without crossing the do-not-invent line — personal names, D-9, region, and revisit-order stay flagged for Josh.

Why this option. Approve-with-amendments beats Approve-as-is because the roster fails its OWN binding executable rule in three spots, and those defects are cheap to leave in but cheaper to fix now: an audit at population time (or in the cascade's Pillar-10 pass) would flag Ch 04-05 strike/strike, Ch 43-44 infiltrate/infiltrate, and the Ch 37-38 attune/attune collision as pillar violations [SRC: CVD L232,L664]. Each fix is a promotion of a secondary verb the row already contains, grounded in the chapter's own T4 canon — no new design, no structural churn. It beats Rework because nothing structural is wrong: I verified the anchor set, the Ch 58 correction, the environmental budget (9 full + 2 staircase = confirmed by count), the Ch 40-41 sole-exception claim (verbatim in T4 L229), and the Pillar 3/8 compliance directly against source, and they all pass. Rework would throw away a verified-winning design to fix three labels. The amendment path is the minimal correct action.

Three-layer fit.

• Surface (L1): L1 is fully served and this is the roster's strongest axis: every chapter (incl. the protected/relief ones) has a real, learnable, fair playable boss with phase escalation that evolves the moveset via telegraph-tightening/multiplication (E33) and never swaps it, >=2 counters per signature, and a Hades reward on both win and death. This directly satisfies CVD's "combat absence is creatively fatal" mandate and the no-pure-DPS-check rule [SRC: CVD L183; SRC: COMBAT L207].
• Mid (L2): L2 is architected as the Preparation axis itself — the investigation layer (sources, ciphers, primary texts, the 42 Confessions found by playing L2) literally becomes ammunition/fairness answers, peaking at Ch 47 and Ch 69. This matches Pillar 5's link-reward loop and the journal/Grimoire investigative substrate [SRC: CVD L195-197,L200-205], and keeps the Preparation axis non-trivial in every row.
• Deep (L3): L3 routes every win through vril mechanics (recharge-site signature response, the purge lineage, substrate perception) rather than declaration, and the Ch 76-77 row holds the show-don't-tell hard line absolutely (no pre-Ch-76 Grand Sage hint anywhere in the roster). This is the CVD L3 rule — metaphysical-as-operational, surfaced through mechanics, never spoken [SRC: CVD L244-246; SRC: CVD L326-328]. The roster does not force L3 engagement to complete L1, satisfying layer independence [SRC: CVD L244].

Test (consistency). At population time, after applying the amendments, verify: (1) NO two consecutive chapter rows share dominant combat type except the sanctioned Ch 40-41 pair — check each row's dominant (primary) mode against its neighbors, confirming Ch 04-05, Ch 43-44, Ch 37-38 are resolved [SRC: CVD L232/L664; T4 L229]. (2) Every row is non-100/0/0 on the axis split and carries a character-engagement/fairness field (Pillar 8 mandatory) [SRC: CVD L218]. (3) Tier-A anchor rows are exactly Ch 8/23/38/55/65 [SRC: COMBAT L211-215,L495]. (4) Environmental ledger = 9 full landmarks + 2 staircase beats, all six COMBAT L349-356 documented instances present [SRC: COMBAT L349-356]. (5) Re-verify all T4 line citations at/after former ~L389 for the post-7e23381 off-by-one caveat. _Expected:_ All five checks pass; the three consecutive-type pairs are gone; anchors and env-budget match canon; T4 line refs re-confirmed.

Adversarial review — HOLDS. I independently read every load-bearing anchor rather than trusting the recommendation, and pulled the actual staged rows from docs/proposals/boss_roster_rows.json. The core recommendation SURVIVES: "Approve-with-amendments" correctly beats "Approve-as-is" and "Rework." But two of the amendments have real flaws — one is a self-contradicting fix, one leans against the authority hierarchy — so the recommendation survives only with corrections, not as written.

WHAT HELD (verified against source):

• Pillar 8 blended/never-single-mode + Ch 17 three-phase template as canonical structure: CVD L214-218, COMBAT L207/L372. Confirmed.
• Pillar 3 hard-line touchability, "combat absence is creatively fatal," Velheim/astral routing: CVD L181-183. Confirmed verbatim.
• Pillar 10 executable test "two consecutive chapters with identical combat feel = pillar violation" (L232) and the "two consecutive chapters never share dominant combat type" rule (CVD L664 AND L879, §16.2). Confirmed verbatim in two places.
• §12.3 environmental boss (perception/vril-management/restraint, victory not measured in damage): CVD L668-676. Confirmed. COMBAT 4-step protocol + failure states + documented instances: L336-356. Confirmed.
• Five Tier-A anchors = Ch 8 / 23-or-25 / 38 / 55 / 65: COMBAT L211-215 AND the doc's own summary L495. Confirmed. The deleted "Ch 58 fifth anchor" correction is RIGHT — T4 L389 shows Ch 58 = first direct engagement, Cassius absent, no fifth-anchor sentence; Ch 58 is correctly staged as a NON-anchor act-climax contingent on ledger D-9.
• Ch 40-41 as the SOLE sanctioned same-type exception: T4 L229 verbatim ("the canonical single exception").
• The three consecutive-type collisions are REAL in the JSON: Ch 04/05 both primary=strike; Ch 43/44 both primary=infiltrate; Ch 37/38 both primary=attune. The claimed secondaries exist (Ch 04 perceive, Ch 05 decode, Ch 43 expose).

WHAT BROKE (two amendment defects):

(1) The Ch 37-38 fix is defective and contradicts its own test. The recommendation proposes a cosmetic dominant_mode LABEL ("no verb change needed"). But CVD L664/L879's rule is about dominant combat TYPE, and BOTH rows are environmental-kind, attune-primary, restraint->purgation wins — the same dominant type. A descriptive label does not change the type, so the Phase-0 pacing audit (and the recommendation's OWN test #1, which demands "NO two consecutive rows share dominant combat type... Ch 37-38 resolved") would STILL flag the pair. The roster's own philosophy declares Ch 36-37-38 "ONE three-beat staircase" (perceive->attune->attune) — an intentional teach-then-test doubling structurally like the Ch 40-41 exception, yet T4 L229 says Ch 40-41 is the ONLY sanctioned exception. So the collision genuinely needs either a real type change (promote Ch 37 restraint/perceive to shift its dominant type, matching the 04-05 treatment) OR a Josh ruling to canonize the staircase as a second exception (which touches T4 L229 = Josh-exclusive). The label-only fix is neither. This is the recommendation's weakest link.

(2) The Ch 57 fauna resolution leans against the authority hierarchy. CVD §17.1 L933 is authoritative and explicit but bans ONLY antagonist-family presence and combat against community representatives — it does NOT ban fauna combat, and says "combat arrives through environmental and astral mode." COMBAT — the Tier-1 doc that OWNS the combat-type domain — permits Ch 57 "person-to-animal with documented dangerous fauna" in TWO places (L371 and the canonical summary L495). T4 L387 "no fauna combat" is the lone dissenter and is tangential to T4's actual domain (family operations). The recommendation recommends recording the no-fauna reading AND flagging COMBAT L371 for a reconciliation EDIT — i.e., editing DOWN a Tier-1 domain doc (twice-stated) to match one sentence in the antagonist doc, on a question CVD does not decide. That inverts CVD > T1-domain-owner. The saving grace: the staged row needs NEITHER fauna nor representative combat (astral/environmental only), so it is compliant under either reading and the flag is genuinely design-neutral. The right move is to KEEP this as a Josh-exclusive reconciliation flag WITHOUT pre-recommending the COMBAT edit direction.

NEITHER defect flips the top-line verdict — the roster passes every load-bearing pillar/anchor check and the only defects are row-level and fixable in place, so Rework is still unwarranted and Approve-as-is still leaves real violations. Disposition await_josh_confirm is correct: the roster does not invent CVD-reserved canon (names, D-9, region labels, revisit-order all stay flagged), and nothing is applied unilaterally.

Canon-faithfulness flags raised:

• Authority-hierarchy lean (Ch 57 fauna): the recommendation would have Josh edit COMBAT L371 (Tier-1 combat-domain doc, which also restates fauna-combat permission at its canonical summary L495) to match T4 L387's lone 'no fauna combat' sentence, on a point CVD §17.1 (L933) does not actually decide. CVD > T1-domain-owner means the combat doc owns which combat TYPES a chapter carries; the recommended direction inverts that. Not a hard-line violation (the staged row is compliant either way), but the recommended RESOLUTION direction is hierarchy-inconsistent and should be downgraded from 'recommend the no-fauna reading + flag COMBAT for edit' to 'keep as neutral Josh-exclusive reconciliation flag, no pre-recommended edit.'
• Self-contradicting amendment (Ch 37-38): the proposed label-only fix (dominant_mode field, 'no verb change needed') does not change the dominant combat TYPE that CVD L664/L879 audits, so it would not satisfy the recommendation's OWN test #1. The roster's philosophy declares Ch 36-37-38 an intentional attune-attune staircase, which collides with T4 L229's 'single exception' (Ch 40-41) — a real conflict a cosmetic label cannot close.

Stronger option the adversary surfaced: For the Ch 37-38 collision specifically (not the whole roster), a stronger option than the cosmetic label is to formally ask Josh to canonize the Ch 36-37-38 teach-then-test staircase as a SECOND sanctioned same-type exception — parallel to the Ch 40-41 doubling — since the roster's own philosophy already treats it as one intentional three-beat structure into the Midpoint anchor. That is cleaner than forcing an artificial type change on a deliberately-designed staircase, but it touches T4 L229 ('single exception') and is Josh-exclusive canon, so it must be raised as a flag, not applied. The recommendation missed this framing entirely and instead papered over the collision.

Corrections to fold in:

• Ch 37-38 amendment: replace the cosmetic dominant_mode label with a REAL dominant-type change on Ch 37 (e.g., promote its restraint/perception element so Ch 37's dominant combat type differs from Ch 38's environmental-attune-restraint), matching the type-changing logic used for Ch 04-05 and Ch 43-44 — OR route it to Josh as a request to canonize the Ch 36-37-38 staircase as a second sanctioned same-type exception, which requires amending T4 L229 and is therefore Josh-exclusive. Do NOT claim the pair is 'resolved' by a label that leaves both rows environmental-attune-dominant.
• Ch 57 fauna flag: keep as a neutral Josh-exclusive canon-reconciliation flag. Do NOT pre-recommend the T4 'no-fauna' reading, and do NOT stage a COMBAT L371 edit. Present both readings squarely to Josh, noting that COMBAT (L371 + L495) is the domain owner and permits fauna combat, CVD §17.1 does not ban it, and the row is compliant under either reading. If anything, the more-canon-consistent default is the COMBAT reading (fauna combat permitted but simply unused in this particular design), not editing COMBAT down to T4.
• Ch 43-44 amendment: prefer promoting Ch 43's part-break objective (a direct-combat texture, a clean type change from Ch 44's stealth) over promoting 'expose' — 'expose' is a social/investigation verb, not one of CVD L654-662's combat-mode categories, so an expose-dominant Ch 43 leaves the dominant COMBAT type ambiguous. The recommendation lists both; make the part-break the primary path so the fix is a genuine type change.

Apply actions. On Josh's sign-off, apply in docs/proposals/boss_roster_rows.json (proposal file only; T0_Boss_Encounter population is a separate post-approval step): (1) Ch 04 row — promote the existing secondary \"perceive (find the coercer behind the screen)\" to primary_verb and demote strike to secondary (or, equivalently, do the swap on Ch 05 promoting \"decode\"); adjust axis_split so Puzzle/Preparation leads (e.g. Ch 04 -> 30/45/25) to make the dominant-mode shift real, not cosmetic. Pick ONE of Ch 04/Ch 05 to flip so 04-05 no longer both read melee-dominant. (2) Ch 43 row — promote existing \"expose\" (or the part-break objective) to primary so Ch 43 is not infiltrate-dominant back-to-back with Ch 44; leave Ch 44 as infiltrate. (3) Ch 37/38 rows — add an explicit dominant_mode field (Ch 37 = \"storm-rigging traversal-endurance\", Ch 38 = \"seven-vril-seat protocol exam\") so the attune/attune primary-verb label no longer reads as a shared dominant type; no verb change needed. (4) Ch 57 row — change the fauna item from \"[DECISION NEEDED - JOSH]\" carried-as-design-neutral to a resolved note recording Josh's ruling; recommend recording \"no fauna combat per T4 L387 + CVD §17.1 maximum protection (more-protective reading wins); the COMBAT L371/L495 person-to-animal listing is superseded at Ch 57 only\" and flag COMBAT L371 for a one-line reconciliation edit at cascade time. Leave untouched and still-flagged for Josh: all personal-name TBD stubs (naming canon), Ch 58 D-9 contingency, Ch 43 region label (T4 Appalachia vs WGM Southeast), Ch 59-63 revisit region-order (T4 vs WGM). Do NOT populate T0_Boss_Encounter or normalize chapter_id to CH_NN until a separate ruling authorizes registry population.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Ch 57 fauna combat: T4 L387 "no fauna combat" vs COMBAT L371/L495 "person-to-animal with documented dangerous fauna" — genuine cross-doc contradiction at the maximum-protection chapter; recommend the more-protective T4 reading (CVD §17.1) but the ruling is Josh's, and it implies a one-line COMBAT reconciliation edit.
• [DECISION NEEDED - JOSH] Personal names for all TBD-stub bosses remain Josh-exclusive naming canon (correctly deferred in the roster).
• [DECISION NEEDED - JOSH] Ch 58 non-anchor boss existence is contingent on ledger decision D-9 (docs/archive/LEDGER_BATCH1.md §14) — not resolvable here.
• [DECISION NEEDED - JOSH] Ch 43 region label (T4 Appalachia vs WGM Southeast/Gullah Geechee) and Ch 59-63 revisit region-order (T4 vs WGM) are open canon-placement discrepancies for Josh.

---

2. Approve/amend the 13-row school roster (Candidate A synthesized)

• Key: 0.E-school-roster Type: creative Disposition: await_josh_confirm

Options

• APPROVE with two carve-outs — Ratify all 13 rows as the T0_School_Registry seed and the teach->grant->showcase / verbs-not-numbers / test-or-narrate-never-count design spine as-is, EXCEPT (a) demote the grafted 'no 14th verb family' governance rule to a flagged recommendation (it constrains CVD L189's open clause = future canon), and (b) log the 120-gate 'not in CVD' caveat as resolved-upward (the figure is canonical at T1 tier via Trades_Master, not merely derived). Registry population proceeds after sign-off.
• Canon support: [SRC: CVD L189] names exactly these 13 schools verbatim (Egyptian Ch15, Mesopotamian Ch17, Levantine Ch18, Taoist Ch21, Eleusinian Ch23, Druidic Ch24, Hermetic Ch27, Sami Ch29, Mesoamerican Ch47, + threshold Ch66/67/68/69) - the roster invents no school and fills no more. [SRC: CVD L190] executable test 'School technique that does not run on vril mechanics = pillar violation' is satisfied per row. [SRC: CVD L604-646/12.1] tier ladder T1-2 verb / T3-5 modifier / T6-8 expression + Portal Grade at T5 + Inner-Order at T6 matches the roster's tier discipline. [SRC: CVD L977/17.5] 'Every tier requires a quest-unlock' grounds test-or-narrate-never-count. [SRC: TRADES_Master L99] independently derives the '120-gate roster - ten trades x twelve quest gates per trade' at T1 tier.
• APPROVE as-is (roster author's package unchanged) — Ratify the entire synthesized package including the governance rule as binding and the 120-flag as an open caveat, exactly as written. Fastest path; registry population immediately.
• Canon support: Same L189/L190/12.1/L977 support as Option 1. The difference is this option treats the grafted governance rule ('never mint a 14th verb family') and the Poison-boundary exclusion as already-ratified rather than flagged. [SRC: CVD L189] open clause 'Additional regional schools emerge per research deliverable' is authored-over by the governance rule under this option - i.e. it silently narrows a canon clause Josh has not ruled on.
• AMEND - repair the tier-count arithmetic first — Approve the 13 rows but require the roster to reconcile the '12 tiers x 10 trades = 120' framing against the named-ten-tiers reality (Known-issue-2) before registry population.
• Canon support: [SRC: CVD L604] '12.1 Unified mastery progression - Twelve tiers' vs the ten names Neophyte..Magus enumerated L614-646. [SRC: TRADES_Master L99] uses 'twelve quest gates per trade' to reach 120. [SRC: Known-issue-2, CLAUDE.md] Josh's standing ruling: preserve the twelve-tier label verbatim, flag only, never auto-correct - which this option honors by flagging but blocks population until Josh rules the gate math.
• REWORK - re-derive gate allocation bottom-up — Reject the top-down '~2-3 gates per school of 120' allocation and require a per-trade x per-tier gate ledger before approval.
• Canon support: [SRC: TRADES_Master L99] the 120 is 'ten trades x twelve quest gates per trade' - a closed grid. The roster assigns approximate shares, not the closed grid. [SRC: CVD L191 cross-ref] points gate detail to T0_School_Registry + per-chapter design, i.e. downstream of this approval - so a closed grid is not required before approval.

Recommendation — APPROVE with two carve-outs

How it fits the CVD. The roster is a faithful instantiation of CVD Pillar 4, not an invention. [SRC: CVD L189] enumerates the 13 schools verbatim; the roster fills exactly those 13 and mints none. Pillar 4 design intent 'Schools teach techniques that run on vril mechanics' [SRC: CVD L188] maps onto teach->grant->showcase, and the L190 executable test ('technique that does not run on vril = violation') is satisfied per row by citing a vril-consuming ABIL ability (Levantine COMBINE = Layer B dual-base cast consuming vril from both pools [SRC: ABIL L897]; Maatherion DECLARE = True Voice substrate-frequency speech [SRC: ABIL L777]). The tier discipline (T1-2 new verb / T3-5 modifiers / T6-8 combination, never numbers) is the shape of 12.1 [SRC: CVD L604-646]: Outer-Order T1-5 with Portal-Grade wisdom-test at T5 [SRC: CVD L626-628], Inner-Order at T6 where 'Substrate perception begins ... earned through a canonical Adept-initiation questline - NOT just hitting numerical threshold' [SRC: CVD L632-633]. Test-or-narrate-never-count is the executable form of 17.5 L977 'Every tier requires a quest-unlock.' The Maatherion capstone (weapons unresponsive, perceive+pacify win, 42 Confessions as recitation) is exact canon [SRC: CVD L175-177, L676; VRIL 12.3 L390]. The Sami max-care school routes its exam through the 4-step environmental protocol with no practitioner-combat, honoring 17.1 care and Pillar 3's C-refined framing [SRC: CVD L923, L929; T1_Combat L338-341].

How it builds canon. Approval stands up the empty flagship T0_School_Registry (currently a DRAFT v0.1 stub with 0 rows - the exact CANON_AUDIT Known-issue-4 gap) with 13 canon-anchored rows, and discharges Josh's Ruling 2 (Claude CREATES adversarially-reviewed school rosters). Forward it: (a) gives the Phase-6 spine cascade a per-chapter school-verb-grant map so /spine-chapter can populate School/Gate fields without re-deriving; (b) supplies Pillar-4 executable-test fixtures (every row names its vril-consuming ability, so the harness can assert L190 mechanically); (c) closes the orthogonality question the endgame needs - each row's endgame_consumer proves the verb stays load-bearing at Ch 70-77, which the Ch 70-77 hook plan then consumes; (d) surfaces two real downstream decisions cleanly (Freedom-Form/Collective-Surge redesign candidates; the twelve/ten tier-count reconciliation the 120-gate math inherits) rather than burying them.

Why this option. Option 1 beats APPROVE-AS-IS because the package as written silently narrows a live canon clause: CVD L189 ends 'Additional regional schools emerge per research deliverable' [SRC: CVD L189], and the grafted governance rule ('never mint a fourteenth family') authors a constraint on that open clause - exactly the cardinal-rule line, so it must be flagged not default-ratified. Option 1 also corrects the one over-conservative flag: the roster carries '120 is CLAUDE.md-derived, not in CVD', but TRADES_Master L99 derives and names the '120-gate roster' at T1 domain-doc tier [SRC: TRADES_Master L99] - the figure is canonical above CLAUDE.md, so that caveat resolves upward. Option 3 over-blocks: Josh's Known-issue-2 ruling is preserve-and-flag never auto-correct, and 120 holds under either count (12 gates x 10 trades), so gating population on it contradicts the standing ruling. Option 4 demands a closed 120-cell grid that canon explicitly defers downstream to T0_School_Registry + per-chapter design [SRC: CVD L191], front-loading cascade work for no fidelity now. Both adversarial judge lenses already selected Candidate A unanimously, and my independent spot-checks confirmed the load-bearing anchors (L189 school list, L190 vril test, 12.1 tier ladder, L977 quest-gate rule, L901 Layer-B guardrail, Freedom-Form scalar at L749, zero 'school' matches in ABIL) - the substance is sound; only the two carve-outs need Josh's eye.

Three-layer fit.

• Surface (L1): L1 is preserved in every row: each school questline is playable action, not a lecture - training duels, the environmental-protocol exam (Sami Ch 29 drum-site), combat-puzzle climaxes (Pohljar Sampo contest), and the Maatherion materialized-argument fight where weapons-unresponsive is canonical yet the full non-weapon verb catalog keeps it action-packed [SRC: CVD L175-177, L676; T1_Combat L338-341]. Verbs-not-numbers directly serves L1: a new verb changes what the player DOES, so surface combat stays fresh across 77 chapters without stat inflation (Pillar 10/11 anti-staleness).
• Mid (L2): L2 is the school's documented real tradition as investigatable substrate - Golden Dawn grades (Hermetic), Kalevala Sampo (Pohljar), Enuma Elish (Tiamarath), Ma'at 42 Confessions (Maatherion), Eleusinian initiation (Greek) - referenced per 17.1 care tiers, never reproduced where protected [SRC: CVD L189, L929]. Each school is a Grimoire shelf the depth player reads behind the verb they wield.
• Deep (L3): L3 is Pillar 4's own executable test made literal: every technique runs on vril mechanics [SRC: CVD L190], so the metaphysical is operational, never declared. The threshold-school relay (READ-SUBSTRATE -> CONFIGURE -> COMMUNE -> DECLARE, Ch 66-69) surfaces the convergence through Pattern - the same verb-grammar recurring across disconnected traditions - feeding the Ch 76-77 single canonical surfacing with no pre-Ch-76 hint [SRC: CVD L1032-1046].

Test (consistency). Before registry population verify the two carve-outs and re-confirm load-bearing anchors: (1) CVD L189 school list == the 13 rows (confirmed exact match); (2) each row's granted verb cites a vril-consuming ABIL ability to satisfy L190 (spot-checked Levantine ABIL L897, Maatherion ABIL L777, Egyptian ABIL L809 - all vril-costed); (3) confirm TRADES_Master L99 grounds the 120-gate figure at T1 tier, record the flag resolved-upward; (4) confirm the governance rule is demoted to a flagged recommendation not written into the registry as binding, because CVD L189's open clause is unruled; (5) after population run the Pillar-4 gate: assert every School_Registry row has a non-empty vril-ability anchor. Known-issue-2 (twelve label / ten names) stays preserved verbatim, flagged only. _Expected:_ L189 list matches 13/13 (confirmed); all 13 verbs vril-anchored; 120 figure attributed to TRADES_Master L99 (T1) not just CLAUDE.md; governance rule and Poison-boundary carried as flags; Pillar-4 gate passes on the populated registry with zero rows lacking a vril anchor.

Adversarial review — HOLDS. I independently re-read every load-bearing anchor rather than trusting the recommendation's citations; all held.

VERIFIED: (1) CVD L189 names exactly the 13 schools verbatim and carries the open clause "Additional regional schools emerge per research deliverable" — the roster fills exactly 13 and invents none. (2) TRADES_Master §4 (~L99) states verbatim "the canonical 120-gate roster — ten trades multiplied by twelve quest gates per trade," at T1 domain-authority tier — carve-out (b) legitimately resolves the 120 flag upward. (3) CVD L175-177 confirms "materialized-argument combat entities" and "weapons unresponsive to input, canonical"; VRIL 12.3 L390 confirms the 42-Confessions recitation as the Ch 69 Maatherion Truly-Good resolution mechanic — the Maatherion capstone is exact canon, and the row's L676 cite is used correctly (only for "victory not measured in damage"). (4) COMBAT L338-341 has the exact 4-step Perception/Vril-management/Restraint/Purgation protocol with L345-347 graduated-consequence failure states; Sami Ch 29 is a listed instance. (5) grep for "school" in ABIL = 0 matches, confirmed — institutional anchors correctly draw from CVD L189, not ABIL. (6) True Voice (ABIL L777) is vril-costed, so the DECLARE verb runs on vril per the Pillar-4 executable test (CVD L190).

REFUTATION ATTEMPTS THAT FAILED: (a) The governance rule "never mint a fourteenth verb family" DOES constrain CVD L189's open clause = future-canon authorship = Josh-exclusive — but carve-out (a) already catches this and demotes it to a flagged recommendation rather than writing it into the registry as binding. Correct handling of the cardinal rule. (b) Option 2 (APPROVE-AS-IS) would silently ratify that governance graft; the recommendation correctly rejects it. (c) Option 3 over-blocks by gating population on the twelve/ten reconciliation, contradicting Josh's standing preserve-verbatim-never-auto-correct ruling; Option 4 demands a closed 120-cell grid that CVD L191 explicitly defers downstream to T0_School_Registry + per-chapter design. Option 1 is the correct minimal-intervention choice. (d) The verb labels (SYNTHESIZE/COMMUNE/COMBINE/DECLARE) are descriptive names for real vril-costed ABIL abilities (Layer B combos, Vril Communion L819, True Voice L777), not invented mechanics. (e) Disposition is await_josh_confirm with apply_actions gated behind sign-off AND a Pillar-4 vril-anchor test on all 13 rows — appropriately conservative for a flagship-registry stand-up.

NO HARD-LINE / §17 VIOLATION: L1-playable is preserved in every row including the weapons-unresponsive Maatherion fight (routed through the non-weapon verb catalog, consistent with CVD L175-177 materialized-argument combat, honoring Josh's Ruling 1); Sami Ch 29 max-care exam routes through the 4-step protocol with no practitioner-combat (§17.1 / CVD L923); no pre-Ch-76 Grand Sage hint (§17.9). The Poison-boundary and Freedom-Form/Collective-Surge items are correctly carried as flags, not auto-fixed (numerical/ability canon = Josh-exclusive).

THREE-LAYER: all three claims hold and are canon-anchored — L1 playable every row, L2 documented-tradition-as-investigatable-substrate under §17.1 care tiers, L3 Pillar-4 executable test made literal with the Ch 66-69 threshold relay surfacing convergence through Pattern and no pre-Ch-76 hint.

ONE PRECISION NOTE (does not change verdict): carve-out (b)'s phrasing "canonical at T1 tier ... not merely derived" is slightly loose — TRADES §4's 120 is itself a T1-tier DERIVATION from CVD §12.1 × §17.5; CVD never states "120." The flag resolves upward in AUTHORITY (canonical at T1 domain tier) but not INTO CVD text, and the 120 TOTAL stays contingent on the Known-issue-2 twelve/ten reconciliation — which open-subflag 4 correctly preserves. Internally consistent; suggest the ROSTER_PROPOSALS §3 edit say "canonical at T1 tier as a domain-doc derivation (Trades_Master §4), 120 total contingent on the preserved twelve/ten tier-count flag." Also worth surfacing the Hermetic "diegetic doctrine" (CVD L977 as in-fiction Order teaching) to Josh as a third creative graft alongside the governance-rule and Poison-boundary grafts, so all three interpretive additions get his eye.

The recommendation is canon-faithful, the anchors check out, and no material flaw changes the verdict. Disposition await_josh_confirm is correct.

Corrections to fold in:

• Precision on carve-out (b): TRADES_Master §4's '120-gate roster' is itself a T1-tier DERIVATION from CVD §12.1 (twelve tiers) × §17.5 (per-tier quest-unlock); CVD never states '120.' The flag resolves upward in AUTHORITY (canonical at T1 domain tier) but not INTO CVD text. Recommend the ROSTER_PROPOSALS §3 edit read 'canonical at T1 tier as a domain-doc derivation (Trades_Master §4), with the 120 total contingent on the preserved twelve/ten tier-count flag' rather than implying an independent CVD figure — open-subflag 4 already preserves the residual contingency, so this is a wording tightening, not a substantive reversal.
• Surface the Hermetic 'diegetic doctrine' graft (CVD L977 reframed as in-fiction Order teaching) to Josh as a third creative graft alongside the governance-rule and Poison-boundary grafts. It is harmless (interpretive flavor on a canonical rule, no mechanical change) but should get Josh's eye rather than only the two carve-outs the recommendation foregrounds.

Apply actions. On Josh's sign-off: (1) Populate registries/"T0_School_Registry [DRAFT v0.1]" with the 13 rows from docs/proposals/school_roster_rows.json (fields: school_name, chapter_home, verb_granted, tier_structure, quest_shape, showcase_chapter, gates, endgame_consumer, anchors); flip DRAFT v0.1 -> ACTIVE v1.0 only after the Pillar-4 vril-anchor gate passes on all 13 rows. (2) In docs/proposals/ROSTER_PROPOSALS.md §3 edit the audit's 120-gate flag: change 'the 120 phrase is NOT in CVD - flagged, not invented' to note it IS canonical at T1 tier via T1_Trades_Master L99 ('ten trades x twelve quest gates per trade'), resolving the caveat upward while preserving the Known-issue-2 twelve/ten note. (3) In school_roster_rows.json philosophy + graft_log demote the 'never mint a fourteenth verb family' governance rule from binding to [FLAGGED RECOMMENDATION - constrains CVD L189 open clause; awaits Josh]. (4) Record the open_subflags below in docs/CANON_AUDIT.md 'Josh's rulings' queue. (5) Do NOT auto-redesign Freedom Form (ABIL L749) or Collective Surge (ABIL L827) - carry as flagged ability-tree-pass candidates only. (6) git commit: 'Populate T0_School_Registry (13 rows) from approved 0.E school roster; resolve 120-gate flag upward to Trades_Master L99; demote 14th-family governance rule to flag'.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] GOVERNANCE RULE vs CVD L189 open clause: the grafted rule 'any future school must subdivide an existing verb family or occupy a modifier grade, never mint a fourteenth verb family' constrains CVD L189's 'Additional regional schools emerge per research deliverable.' A constraint on future canon - Josh-exclusive. Recommend adopting (it protects endgame verb-orthogonality per ABIL L280) but must be Josh's ruling, not a default graft.
• [DECISION NEEDED - JOSH] POISON-BOUNDARY EXCLUSION: 'no school teaches Poison; toxicology stays with the Tamil Siddha trade questline' (ABIL L853/L857) is well-supported but decides no future school may claim Poison - a canon exclusion. Recommend adopting; flag for Josh.
• [DECISION NEEDED - JOSH] FREEDOM FORM (ABIL L749, Tier 8) and COLLECTIVE SURGE (ABIL L827, Tier 7) redesign: the audit correctly flags these as pure numeric buffs at expression-tier slots. Whether to redesign them into expression-class verbs in a later ability-tree pass is Josh's call (numerical/ability canon) - never auto-fixed.
• [DECISION NEEDED - JOSH] TIER-COUNT / 120-GATE ARITHMETIC (Known-issue-2 propagation): 12.1 says twelve tiers but names ten; TRADES_Master L99 derives 120 as 'twelve quest gates per trade.' The 120 total holds, but the twelve/ten reconciliation is Josh-exclusive per the standing preserve-verbatim ruling. Flagged, not corrected.

---

3. Ch 70-77 Deepest-Layer Hook Plan (P5 replacement engine)

• Key: 0.E-Ch70-77-deepest-layer-hook Type: creative Disposition: await_josh_confirm

Options

• Approve as written — Ratify ROSTER_PROPOSALS.md §4 verbatim: the three-hook engine (Witnessing/reverse-linking, place-as-text, Convergence choreography) replaces culture-to-culture linking across Ch 70-77 + Epilogue, with the four §4.5 flags kept at their safe defaults.
• Canon support: The engine is built entirely on affirmed anchors: Thread 1 FRAMEWORK-LEVEL CULMINATION 'fifteen-parallel catalog's testimonial content affirmed at pre-decline atmospheric density via direct protagonist experience' [SRC: T22 L1166, L96]; the echo-source survey is a literal Thread 1 anchor 'sacred sites identified as fairy-realm sources of human-world echoes encountered Ch 2-57' [SRC: T22 L1170]; place-as-text is canon ('grief reads in the realm's current state at landscape register'; 'fairy-realm script natively legible at full mastery') [SRC: RCM L6536]; convergence choreography rests on 'all six Major Threads reach their unifying resolution at the Ch 76-77 waterfall' [SRC: T22 L982] and the reveal 'references the cumulative vril-seat-signature pattern accrued across the arc' [SRC: RCM L6874; CVD §19.5]. Pillar 12 mandate that forces the need: 'Chapter without all three layers represented = pillar violation' [SRC: CVD L246].
• Approve with amendments (RECOMMENDED) — Ratify the engine, but (a) tighten one show-don't-tell exposure at §4.2.3/§4.3-Ch76 so the Convergence map never renders strand-count-toward-a-named-point in a way a player could read as a pre-reveal thesis hint; (b) formally mark the Witnessing link-semantics inversion (culture-to-culture -> culture-to-referent) as the one true new-P5-canon seam requiring Josh sign-off, not silently folded in as 'still satisfies the test'; (c) keep flags 1 (battle-myth confirmability) and 2 (serpent Witness slot) at safe default = no new canon.
• Canon support: Same affirmed-anchor foundation as Option 1, PLUS the discipline hooks the plan itself cites but leaves as prose rather than binding gates: Pillar 5 test 'Link-action without scholarly basis = pillar violation. Cross-cultural reward without primary-source documentation = pillar violation' [SRC: CVD L197] — the Witness link keeps the Earth-side primary source as one documented end, but it redefines 'link' from testimony-to-testimony to testimony-to-referent, which is a genuine mechanic extension of a player-facing Pillar; the Resolution framing rule 'must describe the threads' resolutions at framework level without hinting at the specific revelation content' [SRC: T22 L1006] and §17.9 'Pre-Ch 76 Grand Sage identity hints are forbidden' [SRC: CVD L1038] bind the Convergence-map visualization tightly.
• Amend the mechanic name/scope (Witnessing -> pure perception-verb, no 'link' framing) — Keep all three hooks but strip the word/mechanic 'link' from Witnessing entirely: reframe realm-side confirmation as an ordinary L3 perception event (Vril Sight recognition) that enriches an already-Linked node's journal, never as a new Grimoire link-state. Avoids touching Pillar 5's link semantics at all.
• Canon support: Conservative reading of Pillar 5: 'link' is explicitly a culture-to-culture player action [SRC: CVD L196 'Encountering related material in another culture enables a link action']; RCM routes fairy-realm traditions as 'documented cultural-tradition reference, not Frame B uptake' [SRC: RCM L6536], and perception-through-mechanics is the pure §18.2 L3 route [SRC: CVD L1142]. This option keeps every reward denomination (lore + cross-cultural weapon abilities, technique hybrids, intuition tiers) [SRC: CVD L196] but delivers them as perception rewards, not link rewards.

Recommendation — Approve with amendments (RECOMMENDED)

How it fits the CVD. The plan solves a real, canon-created problem: Pillar 12 makes all three layers mandatory every chapter ('Chapter without all three layers represented = pillar violation' [SRC: CVD L246]; per-chapter L2 + L3 required [SRC: CVD §17.9 L1040-1042]), yet canon itself cuts the Pillar-5 fuel line at Ch 70 ('no Earth-cultural primary-source equivalents' [SRC: RCM L6536], repeated Ch 71-77 + Epilogue [SRC: RCM L6642/6682/6720/6842/6874/6906/6968], and Thread anchors thin to one-or-none [SRC: T22 L1166/L1174/L1176]). The replacement engine carries L3 exactly the way §18.2/§18.5 demand — through mechanics and Pattern, never declaration: Witnessing is a perception verb; place-as-text is landscape/architecture reading; Convergence is the Thread Map physically braiding. Crucially it respects the single hardest CVD line here — 'The Grand Sage Ch 76-77 reveal is the single canonical direct surfacing... Pre-Ch 76 hints forbidden' [SRC: CVD L1132, L1038] — by keeping the map at framework level and never labeling the convergence point [SRC: T22 L1006]. It also honors Pillar-12 layer independence ('No layer forces engagement with other layers' [SRC: CVD L244]): every deepest-layer reward is textural/expression-tier and never gates L1 combat, and tier advancement stays quest-gated per the mastery hard line [SRC: CVD L977]. The amendment exists only to make the two thinnest CVD seams explicit rather than absorbed.

How it builds canon. Approving instantiates the Ch 70-77 + Epilogue rows of the mode-variety / three-layer checks that Phase 6 needs anyway (the flagship mode_variety_check registry is 0/79 populated per Known-issue #4), giving the cascade a canon-anchored L2/L3 column set for the eight chapters where it would otherwise have nothing to populate. It converts the P5 engine from 'goes dark for 8 chapters' into a documented continuation contract that runs through the Epilogue and hands off cleanly to the post-Ch-77 Magister Templi prestige path [SRC: RCM L6906; CVD L643]. It stands up three reusable named mechanics (Witnessing / place-as-text / Convergence choreography) that the spine cascade and T0_Cross_Cultural_Link_Graph can key against, and it correctly defers encounter internals to spine + T1_Combat authority [SRC: RCM L6642/6720/6752] so it builds forward without pre-empting Phase 6.

Why this option. Option 1 is substantively correct and every anchor verifies against the actual canon text (I re-read RCM L6536/L6874, T22 L96/L982-1006/L1160-1180, CVD Pillar 5/12/§18/§17.9 directly — the summary did not overstate). But it folds two genuine canon-invention seams into confident prose: (i) it asserts the Witness link 'stays satisfied' against the Pillar-5 test, when in fact it redefines the player-facing meaning of a link (testimony-to-testimony -> testimony-to-referent) — a real extension of a Pillar, which is Josh-exclusive; (ii) the Convergence-map 'watch 22 strands braid toward one point' is the single highest show-don't-tell exposure in the plan and deserves a binding gate, not just a guard sentence. Option 3 over-corrects: stripping 'link' throws away the plan's best idea — that the 69-chapter Grimoire gets a payoff verb — and the reward denominations are canonically link rewards [SRC: CVD L196], so pure-perception reframing loses fidelity to what P5 pays. The amendment path keeps the strong engine, closes the two seams honestly, and leaves flags 1/2 at safe no-new-canon defaults exactly as the plan already proposes.

Three-layer fit.

• Surface (L1): L1 is fully protected and untouched: the reckoning beat, Realm War Ch 71-73, the Cassius multi-phase boss, the three path-distinct Ch 75 encounters, and the reveal cinematic all read complete without one Witness link; deepest-layer rewards are expression-tier/textural and never gate combat [SRC: CVD L244; RCM L6642]. The engine adds depth, subtracts no surface content.
• Mid (L2): L2 gains a concrete investigative shelf per chapter that canon already stocks — the fairy-and-otherworld catalog (Evans-Wentz, Kirk 1691, Eddas, Mabinogion, Tolkien-as-reception) at Ch 70 [SRC: RCM L6536, L6540-6552], battle-tradition comparative nodes Ch 71-73 [SRC: RCM L6642/6682/6720], Plato/Campbell/banking-history at Ch 74 [SRC: RCM L6752] — all routed strictly Frame A Layer 2 'not Frame B uptake', which the plan preserves.
• Deep (L3): L3 is carried by Pattern-through-mechanics exactly as §18.2/§18.5 require: Witnessing is a perception recognition, the attribution proof at Ch 73 assembles 'one personal will drives the infrastructure' before any dialogue says so [SRC: T22 L1172], and the Convergence map surfaces the thesis-as-convergence without ever naming it. The single guard that must hold is the reveal lock (Ch 76-77 the only direct surfacing [SRC: CVD L1132]) — which the amendment hardens into a check.

Test (consistency). Before spine population of Ch 70-77, verify these anchors hold in current canon (they did at review time): (1) RCM L6536 'no Earth-cultural primary-source equivalents' + Component 7/9 clauses; (2) T22 L1166 Thread 1 framework-culmination wording + L1170 echo-source anchor + L96 fifteen-parallel testimonial line; (3) T22 L982-L1006 six-Major-Thread resolution + framing rule 'without hinting at the specific revelation content'; (4) CVD L246 Pillar-12 three-layer test, L244 layer-independence, L1038 pre-Ch-76-hints-forbidden, L1132/L1046 single-reveal lock, L977 quest-unlock mastery rule, L196-197 Pillar-5 link/scholarly-basis test. AMENDMENT-SPECIFIC gate: confirm the Convergence-map spec never renders a labeled convergence target or a countdown-to-thesis before Ch 76 (show-don't-tell §17.9 check), and that Witnessing rewards are expression-tier only (no tier advancement, no numeric buff) per L977. _Expected:_ All anchors present and unchanged; Convergence-map spec framework-level only; Witnessing rewards carry zero tier-gate and zero numeric buff. Any drift here (esp. a changed reveal lock or a P5-test rewording) re-opens the decision.

Adversarial review — HOLDS. I re-read the cited canon independently rather than trusting the anchors, and the recommendation holds. VERIFIED ANCHORS (all accurate, none overstated): CVD Pillar 5 L195-197 — link = "encountering related material in ANOTHER CULTURE enables a link action"; scholarly-basis + primary-source executable tests exact. CVD Pillar 12 L242-247 — "Chapter without all three layers represented = pillar violation"; "No layer forces engagement with other layers"; L3-declared-in-dialogue = violation. §17.9 L1036-1046 — thesis never declared, pre-Ch-76 Grand Sage hints forbidden, per-chapter L2 + L3 required, single locked reveal. §17.5 L977 — "Every tier requires a quest-unlock." RCM L6536 — verbatim "no Earth-cultural primary-source equivalents," Component 7 strongest presence, Component 9 perception "foreshadowing Ch 76-77... without surfacing it," show-don't-tell absolute, Frame A Layer 2 "not Frame B uptake." RCM L6874 — reveal "references the cumulative vril-seat-signature pattern accrued across the arc per §19.5," phrasing lock owned upstream. T22 L96/L1166 fifteen-parallel testimonial culmination; L1170 echo-source is correctly a "candidate anchor" (plan does not overclaim it as affirmed); L982-1006 six-Major-Thread unifying resolution + the framing rule "without hinting at the specific revelation content" verbatim. The problem the plan solves is real and canon-created: Pillar 12 mandates L2/L3 every chapter, yet canon cuts the Pillar-5 Earth-culture fuel line at Ch 70. REFUTATION ATTEMPTS THAT FAILED: (1) Invented canon? The one genuine invention — Witnessing redefining "link" from culture-to-culture to culture-to-referent — is NOT smuggled in; the recommended Option 2 explicitly extracts it as a Josh-exclusive new-canon seam requiring sign-off, which is the cardinal-rule-faithful move. (2) Authority-hierarchy error? None — CVD wins throughout; no T-doc or draft is treated as binding; §4 still carries [DRAFT for Josh review] (verified in file), nothing pre-approved. (3) Hard-line smoothed over? The opposite — the reviewer caught the Convergence-map "22 strands braid toward one point" as the highest §17.9 exposure and hardened the framework-level-only gate from prose into a binding amendment (supported by T22 L1006 + CVD L1038). (4) Stronger option missed? Considered "defer entire engine to Phase 6" — weaker, because the cascade still needs a canon-anchored engine to populate the mandatory L2/L3 rows; the plan already defers encounter internals and only stands up the engagement layer. (5) Three-layer claim? Holds — L1 untouched, rewards are expression-tier/textural per L977 + L244, both verified; three_layer section is accurate. (6) Disposition too aggressive? No — await_josh_confirm with apply_actions touching only docs/proposals (tag flip + two gate lines), explicitly no _source edit until the cascade. That is correctly conservative for a package containing a Josh-exclusive Pillar extension. Minor non-material note: T22 L992 anchors the identity hard line to CVD §15 while the plan cites §17.9/§18; both clauses exist, so this is an additional supporting anchor, not an error. WHAT HELD: the entire engine rests on affirmed anchors; the two thinnest seams (link-semantics inversion, Convergence-map exposure) are exactly the two the amendment path makes explicit. WHAT BROKE: nothing material.

Stronger option the adversary surfaced: none — the option set is adequate. The one alternative not enumerated (defer the entire engine to Phase 6, approve nothing now) is weaker, not stronger: Pillar 12 L246 still mandates per-chapter L2/L3 for Ch 70-77, and the cascade would face the identical gap with no canon-anchored proposal to populate against. The plan already defers encounter internals to spine + T1_Combat authority and only ratifies the engagement layer, which is the correct scope boundary.

Corrections to fold in:

• Non-material: the recommendation cites the pre-Ch-76 identity hard line via CVD §17.9 L1038 and §18; T22 L992 additionally anchors it to CVD §15 ('never hint at the Grand Sage's identity before Ch 76-77'). Both are real; the §15 anchor could be added to the Convergence-map gate line for completeness, but its absence is not a defect.

Apply actions. On Josh sign-off: (1) In docs/proposals/ROSTER_PROPOSALS.md §4, flip the §4 header tag from '[DRAFT for Josh review]' to '[APPROVED - JOSH 2026-07-04]' and add a one-line ruling note recording the two amendments. (2) Add to §4.2.1: bind the Witness link as an approved NEW Pillar-5 mechanic-extension (link semantics = documented-Earth-source -> direct-realm-referent), with the explicit rule 'one end of every Witness link is the node's existing primary source; realm-side is observation, never a new unsourced claim' promoted from prose to a stated executable rule mirroring CVD L197. (3) Add to §4.2.3 and the Ch 76 entry a binding show-don't-tell gate line: 'The Convergence map renders strand progress at framework level only; it never labels, names, or counts-down-to the convergence point before the Ch 76 beat fires [SRC: T22 L1006; CVD §17.9 L1038].' (4) Leave §4.5 flags 1 and 2 at their stated safe defaults (battle-myth = comparative-tier only; serpent Witness slot reserved, no content authored) — record both as 'default accepted, no new canon.' (5) Do NOT populate any registry (mode_variety_check / Cross_Cultural_Link_Graph Ch 70-77 rows) in this step; that happens in the Phase 6 cascade under the consistency gate above. All edits are in-repo, docs/proposals only; no _source canon file is touched until the cascade.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] NEW CANON (Josh-exclusive): the Witnessing mechanic extends Pillar 5's link semantics from culture-to-culture to culture-to-referent. My best-supported recommendation is APPROVE as an extension (one documented Earth-side end preserves the scholarly-basis test [SRC: CVD L197]), but this is a player-facing Pillar change and is Josh's to ratify — never silently folded in.
• [DECISION NEEDED - JOSH] SHOW-DON'T-TELL (Josh confirm): the Convergence-map 'braiding toward one point' is the plan's highest §17.9 exposure. Recommend approving only with the framework-level-only gate above; the mechanic must never label/count-down-to the convergence point pre-Ch-76 [SRC: CVD L1038; T22 L1006].
• [DECISION NEEDED - JOSH] Flag 1 (battle-myth confirmability, plan §4.5): NEW canon only if upgraded. Recommend keep DEFAULT (comparative-tier, not Witnessable-as-realm-war-memory) — the plan works without it and the stronger reading is unsupported by current anchors [SRC: RCM L6642].
• [DECISION NEEDED - JOSH] Flag 2 (fairy-realm serpent tradition, T22 L180 bare anchor): content is Josh/Phase-6 canon. Recommend keep DEFAULT (Witness slot reserved, zero content authored).

---

4. Difficulty and Accessibility spec (ROSTER_PROPOSALS §5) — ratify with amendments

• Key: 0.E-5-difficulty-accessibility Type: creative Disposition: await_josh_confirm

Options

• A. Ratify as-is — Adopt §5 verbatim: one canonical difficulty, in-fiction relief valves, the Attunement Settings assist layer + 4 presets, Exemptus Discipline hardest mode with all 7 prunes, and the defeat model — all names and all new-canon items accepted as written.
• Canon support: Whole spec is anchor-consistent. One-difficulty position grounded: play is canonical not cosmetic [SRC: CVD §9.3 L~2050 'appearance shifts are not cosmetic and not optional... canonical']; diegetic modulation already chosen over menu modulation [SRC: VRIL §5.7 L186 'path-specific gameplay difficulty modulation without explicit integrity-gating']; luck bounded by hard line [SRC: CVD §17.13 L1091-1099]; macro pacing authored not selected [SRC: CVD §17.10 L1052-1062]. Assist-layer integrity-neutrality mirrors canonical respec framing [SRC: ABIL §5B.1 L178 'integrity-neutral by canonical framing']. But it accepts naming/defeat-model/grant-touching prunes as canon, which are Josh-exclusive.
• B. Ratify architecture; hold Josh-exclusive items for explicit sign-off (RECOMMENDED) — Adopt the four-part architecture and the entire assist/accessibility layer now (grounded correction + data-model instantiation). Recommend but hold for Josh's word: the placeholder names, the defeat model as new canon, the two grant-touching Exemptus prunes (Ch 38/55 respecs paid-only; passive-maintenance off), and Exemptus availability timing. Land the vril-polarity readability requirement in T1_UI_UX_Spec.
• Canon support: Architecture is fully anchored (see A). The held items are exactly the ones that invent canon or touch a canonical grant: defeat model is self-declared 'new canon' [SRC: ROSTER_PROPOSALS §5.2 L328]; the Ch 38/55 free respecs are canonical grants [SRC: ABIL §5B.1 L180], and passive-maintenance mitigation is canonical [SRC: COMBAT 'Mitigation']; names are naming canon [SRC: CVD §17.7 L999-1014, Josh-exclusive]. CLAUDE.md cardinal rule: do NOT invent canon — flag with a proposed resolution.
• C. Offer a menu-difficulty option (Easy/Normal/Hard) — Reject the one-difficulty position; add a conventional difficulty menu that scales HP/damage.
• Canon support: Directly contradicted by canon. A menu forks a canonical biography into non-canonical variants [SRC: CVD §9.3 L~2050; §17.4 L963 'Integrity is a biography parameter, not a stat... canonically operational']. HP/damage scaling is the SotE/Wukong failure the binding design principles reject [SRC: docs/ROADMAP.md §0.E]. No canon support; listed only to be dismissed.

Recommendation — B. Ratify architecture; hold Josh-exclusive items for explicit sign-off

How it fits the CVD. The spec is an unusually faithful extension of existing CVD philosophy rather than an invention on top of it. (1) It honors integrity-as-biography: every assist is integrity-neutral and no toggle touches a moral/world-state consequence, mirroring the canonical respec framing [SRC: ABIL §5B.1 L178; CVD §17.4 L963]. (2) It honors show-don't-tell: hints route through the journal voice (Pillar 6), so reaching for help deepens Layer 2 contact rather than bypassing it [SRC: CVD §16.3 Pillar 6 L200; §17.9 L1042]; the defeat model is a substrate event (consciousness withdrawal to an attuned recharge source), keeping Layer 3 real through mechanics [SRC: VRIL §1; §5; CVD §17.9 L1040]. (3) It honors the hard lines absolutely: no toggle creates combat against protected community representatives [SRC: CVD §17.1 L933], skips an environmental-protocol step, restores protected-site recharge to negative signatures [SRC: VRIL §5.7 L186], or touches the sigma constraint [SRC: CVD §17.13]. (4) It preserves the three finales as canonical and equally-weighted, refusing to let any assist neutralize Evil-path friction [SRC: CVD §7.2 L318-324; §17.4 L969]. (5) The vril-polarity readability floor is a genuine L1-playability requirement under Josh's Ruling 1: the environmental-boss perception verb is unplayable for colorblind players if polarity is color-only [SRC: VRIL §1 polarized states; CVD §17.9 L1040 per-chapter L3 surfacing]. The only places it exceeds canon are the four flagged Josh-exclusive items, which option B holds.

How it builds canon. Ratifying establishes the game's first difficulty/accessibility canon — a section CVD does not currently cover — and does so by codifying choices canon already implies, so it hardens rather than expands the surface. Concretely it: (1) writes a binding tuning contract onto the boss roster rows (>=1 counter per signature, verb-taught-before-tested, no pure DPS checks, E33 telegraph escalation), giving the Phase-6 spine cascade an executable fairness spec for every Ch 2-77 boss [SRC: ROSTER_PROPOSALS §5.1 L306-311]; (2) generalizes the canonical Hades-loop already present at environmental bosses (partial resolution → graduated consequence → matured revisit) into a whole-game defeat-productive model [SRC: COMBAT env-boss failure states; §5.2 L322]; (3) stands up T1_UI_UX_Spec — currently a 2-line empty DRAFT stub — with its first real spec item (multi-channel polarity readability), a Phase-0-hardening win the cascade needs anyway. The held items, once Josh rules, extend canon deliberately (a named checkpoint verb, a named prestige-difficulty mode) rather than by drift.

Why this option. B is the maximum-ratifiable split: adopt every grounded, gate-safe, hard-line-respecting element now; route only genuine canon-invention (names, defeat model, grant-touching prunes) to Josh, exactly as the proposal's own DECISIONS 12-17 request.

Three-layer fit.

• Surface (L1): L1: assists exist precisely to keep L1 action-packed and playable in every chapter for every player (Josh Ruling 1), and Exemptus deepens L1 for experts — neither adds nor removes one piece of content [SRC: ROSTER_PROPOSALS §5.5 L398]. Outgoing damage is never inflated, so every boss still tests its stated verb at full definition [SRC: §5.3 L366]. The vril-polarity multi-channel floor is a hard L1-playability requirement: color-only polarity makes the environmental-boss perception verb unplayable for colorblind players [SRC: VRIL §1; CVD §17.9 L1040].
• Mid (L2): L2: the inscription/puzzle hint system is delivered diegetically in the journal's voice, so a struggling player reaching for help touches investigatable cultural depth on the way — the assist deepens Layer 2 contact rather than bypassing it [SRC: ROSTER_PROPOSALS §5.3 L363; CVD §16.3 Pillar 6 L200]. The defeat-writes-journal-observation mechanic adds L2 content on failure (one recorded telegraph-and-counter per defeat) [SRC: §5.2 L331].
• Deep (L3): L3: the metaphysical stays real through mechanics in every configuration — perception assist amplifies substrate signal rather than replacing the substrate verb, and the defeat model is a substrate event (consciousness withdrawal to an attuned recharge source), not a UI event [SRC: §5.5 L400; VRIL §1, §5; CVD §17.9 L1040 per-chapter L3 surfacing required]. No assist can restore protected-site recharge to negative signatures, so the L3 world-response substrate is untouched [SRC: VRIL §5.7 L186].

Test (consistency). Before any registry/spec write, verify the four held items against canon and confirm the architecture is contradiction-free: (1) no assist/preset/Exemptus prune crosses a §17 hard line — re-check each toggle against CVD §17.1 (protected representatives, L933), VRIL §5.7 L186 (protected-site non-response), CVD §17.13 (sigma), and the four-step environmental protocol in COMBAT §9.2; (2) the defeat model does not contradict any existing respawn/checkpoint canon (grep COMBAT + VRIL for existing death/respawn handling — none should conflict); (3) the two Exemptus grant-prunes name the exact canonical grants they modify (ABIL §5B.1 L180 free respecs; COMBAT 'Mitigation'); (4) Evil-path friction remains un-neutralized by every assist [SRC: CVD §7.2, §17.4 L969]. Expected: zero hard-line crossings; the only canon deltas are the four Josh-flagged items. _Expected:_ Architecture verifies clean against all cited anchors (confirmed this review). Four residual items require Josh's ruling; nothing else blocks.

Adversarial review — HOLDS. I independently re-read every load-bearing anchor rather than trusting the recommendation's citations. They check out with near-total fidelity, and the recommendation's core judgment — ratify the grounded architecture but hold the four Josh-exclusive items, disposition await_josh_confirm — is canon-faithful and correctly conservative.

VERIFIED ANCHORS (all confirmed against readable canon):

• ABIL §5B.1 L178 "respec is integrity-neutral by canonical framing" and L180 free respecs at Ch 38/55/post-77 — EXACT. Confirms both the assist integrity-neutrality mirror AND that Exemptus prune #14 genuinely touches a canonical grant (correctly held for Josh).
• VRIL §5.7 L186: protected-site non-response (Ch 57/29/42 "do not respond to negative-vril signatures at all") + "path-specific gameplay difficulty modulation without explicit integrity-gating" — EXACT. The one-difficulty position and the protected-site assist-exemption both hold.
• VRIL §1 L19: vril "polarized into positive and negative vibrational states" — EXACT. The colorblind/multi-channel L1-playability floor is real, not invented.
• CVD §9.3 L419: "appearance shifts are not cosmetic and not optional. They are canonical" — EXACT. Option C (menu difficulty) is correctly refuted.
• CVD §17.1 L923 protected reps "inviolate"; §17.13 L1091 "±1 sigma" hard line; §17.4 L963/969 "Integrity is a biography parameter, not a stat... canonically operational" + "equally weighted in production scope"; §17.7 L999 naming hard line (Josh-exclusive); §7.3 L328 / §17.9 L1038 Grand Sage locked + pre-Ch-76 hints forbidden — ALL EXACT.
• COMBAT: Mitigation L453 "active lifestyle... counts as passive training" (Exemptus prune #14 target is real); Floor L459 70% floor canon; env-boss L334-347 "Cannot be defeated through damage / partial resolution produces graduated consequence"; World Boss Kill L241 regional floor lowers / high-end unaffected; Hardcore L297 cosmetic integrity-band titles — ALL EXACT.
• Integrity_Paths §8.5 L327-333 path-variable journal voice — EXACT (defeat-model journal texture is properly grounded).
• T1_UI_UX_Spec is a 35-byte DRAFT stub — the "empty stub" claim is literally accurate; landing polarity there awaits the DRAFT→ACTIVE flip (correctly NOT applied now).
• ROADMAP §0.E is Josh's own Ruling-2 authoring frame (Celeste/fairness-checklist/E33/HP-never-inflate), which sits at ops level BELOW CVD. The recommendation uses it only as the authoring substrate and never lets it override CVD — no authority-hierarchy inversion.

REFUTATION ATTEMPTS THAT FAILED:

1. Invented-canon-as-grounded? The recommendation is scrupulous: it separates the anchored architecture from the four self-declared new-canon/grant-touching/naming items and routes all four to Josh. No cardinal-rule violation.

2. Hard-line smoothed over on the damage-taken-reduction toggle? No. It is incoming-only; a Pillar-8 verb-test (§12.4 L217 "No boss = pure DPS check") tests what RESOLVES the boss (counter/perception/puzzle), which incoming mitigation leaves intact. Outgoing is never scaled, so no verb-test is neutered. Holds.

3. Evil-path friction erased by the vril-flow assist? No. VRIL §5.7 + Integrity_Paths §8.2/§8.6 confirm the assist floors only ORDINARY-source response; protected sites, journal, community, and faction consequences are explicitly exempt (proposal L359/L401). Consistent with §17.4 equal-weighting.

4. Disposition too aggressive? The opposite risk was worth checking — and here Option B's LABEL ("adopt architecture now") is actually looser than its OPERATIONAL disposition. Every apply_action is gated "On Josh's confirm," and item (7) forbids any registry write until the ruling lands. So B holds everything for Josh — correctly conservative for a section CVD does not yet cover. await_josh_confirm is the right disposition; it should not be relaxed to apply_now, and there is no element safe to apply unilaterally today (even the polarity floor needs the UI_UX DRAFT→ACTIVE flip).

Net: the recommendation is canon-faithful, its anchors are real, and no material flaw changes the verdict. Disposition unchanged at await_josh_confirm.

Stronger option the adversary surfaced: none — the three options span the real decision space (ratify-all / ratify-architecture-hold-exclusives / add-menu-difficulty). B is the correct maximum-ratifiable split. The only refinement is internal to B, not a missed option: on the apply_actions ORDERING, DECISION 17 (polarity floor) is the single item with the strongest independent grounding (a hard-playability requirement under Ruling 1, anchored to VRIL §1 + CVD §17.9) and the lowest authorial stakes — Josh could confirm it in isolation even if he defers the naming/defeat-model/prune rulings. Worth surfacing to Josh as the one item he can green-light immediately without entangling it in the Josh-exclusive bundle. This sharpens B rather than replacing it.

Corrections to fold in:

• CITATION-PRECISION DRIFT (inherited, non-material but flag before it enters a spec): the proposal and the recommendation repeatedly call the environmental-boss protocol a 'four-step (perception, vril-management, restraint, purgation)' sequence. Canon (COMBAT L334 and CVD §12.3 L670 heading) names THREE steps — perception, vril-management, restraint — with 'purge/purged' as the resolution OUTCOME, not a listed fourth step. When DECISION 13/17 apply-actions eventually write to COMBAT or T1_UI_UX_Spec, do not codify a canonical 'four-step protocol'; use the canonical three-step framing or explicitly flag the fourth-step addition as a proposed extension for Josh.
• MINOR OVERSTATEMENT in three_layer.surface: the recommendation says assists 'never add nor remove one piece of content.' True for the assist toggles; but Exemptus 'Mid-encounter recharge pruned' (proposal L381) SEALS encounter spaces from recharge access, which is a real mechanical delta from default difficulty (it changes what the encounter contains, even if no HP value moves). This is fine as a hardest-mode prune, but the blanket 'removes nothing' phrasing is slightly generous — Exemptus removes ACCESS/answers by design. Does not change the verdict; the prunes are correctly flagged as mode-local and (for the two grant-touching ones) held for Josh.
• The recommendation cites '§12.2 seven combat modes L654-666' — verified: the seven modes are canonical and correctly enumerated. No correction needed there; noting only that the earlier task brief's line hint (L654-666) matches. Keep the citation.

Apply actions. On Josh's confirm, apply in this order. (1) NAMING (DECISION 12): if Josh accepts placeholders or supplies names, do a global find/replace of 'Attunement Settings', 'Full/Steadied/Guided/Open Practice', 'Exemptus Discipline' with the ruled names across ROSTER_PROPOSALS.md §5 and any downstream refs. (2) DEFEAT MODEL (DECISION 13): on 'apply', author the model into canon — recommended home is a new '§ Defeat and Recovery' subsection in T1_Combat_System_Spec (consciousness-withdrawal-to-last-attuned-recharge-source as diegetic checkpoint; defeat-writes-one-journal-observation recording a witnessed telegraph-and-counter; journal-voice texture path-variable per T1_Integrity_Paths §8.5), then run the normalize-doc critic before any DRAFT→ACTIVE. (3) EXEMPTUS GRANT-PRUNES (DECISION 14): on 'apply', add an Exemptus-mode clause to ABIL §5B.1/§5B.2 (Ch 38/55 free respecs become paid-only in-mode; post-77 unlimited retained) and to COMBAT 'Mitigation' (passive-maintenance off in-mode; 70% decay floor stays). (4) TIMING (DECISION 15): record Josh's ruling on Exemptus availability (proposal recommends post-first-completion with a from-start override). (5) POLARITY (DECISION 17): land the multi-channel (shape/motion/audio alongside color) polarity-readability requirement as the first real spec item in T1_UI_UX_Spec [DRAFT v0.1] (currently an empty stub) — mark it an L1-playability requirement under Ruling 1. (6) TIER-C (DECISION 16): record the ruling (proposal recommends assists client-local in Tier C; Exemptus Tier A/B only) into the eventual spec. (7) Commit with the ruling recorded in the message; do NOT populate any registry until the ruling lands.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] DECISION 12 — naming canon (Attunement Settings / the four Practice presets / Exemptus Discipline): Josh-exclusive per CVD §17.7. Recommend accepting the placeholders as-is; they draw on the game's own practitioner register (Exemptus = Adeptus Exemptus tier per CVD §12.1) but the ruling is Josh's.
• [DECISION NEEDED - JOSH] DECISION 13 — defeat model as NEW canon (consciousness withdrawal to last attuned recharge source + defeat-writes-journal-observation, path-variable voice per T1_Integrity_Paths §8.5): recommend ratify — it is substrate-consistent [SRC: VRIL §1, §5] and makes the checkpoint diegetic — but it is self-declared new canon [SRC: §5.2 L328], so Josh-exclusive.
• [DECISION NEEDED - JOSH] DECISION 14 — two Exemptus prunes that touch canonical grants (Ch 38/55 free respecs → paid-only in-mode; passive-maintenance mitigation off in-mode): recommend ratify as mode-local (canonical grants preserved in default difficulty), but each adjusts a canonical grant [SRC: ABIL §5B.1 L180; COMBAT 'Mitigation'], so Josh-exclusive.
• [DECISION NEEDED - JOSH] DECISION 15 — Exemptus availability timing (from-first-boot vs post-first-completion): recommend the proposal's post-completion default with a from-start override one menu deeper; low-stakes but authorial.
• [DECISION NEEDED - JOSH] DECISION 16 — Tier-C assist locality + Exemptus/world-boss-floor coexistence: recommend assists client-local in Tier C and Exemptus Tier A/B only; needs Josh's ruling on MMO interaction.
• [DECISION NEEDED - JOSH] DECISION 17 — vril-polarity multi-channel readability as a T1_UI_UX_Spec item: recommend confirm as an L1-playability requirement under Ruling 1; the strongest-support item of the set, but formal spec-landing awaits the DRAFT→ACTIVE flip.

---

5. Approve the 3-cell topography re-pacing + rule its 4 carried decisions

• Key: 0.E-TOPOGRAPHY-REPACING-APPROVAL Type: creative Disposition: await_josh_confirm

Options

• A — Approve all three edits (CH_37→vril_management, CH_40→stealth, CH_72→environmental) + remove the 4 waivers; rule the 4 carried decisions per the recommendations — Adopt the SELECT+RECOMMEND package verbatim. Three dominant_combat_mode cells change, zero intensity cells, zero boss-roster rows. The four R1/R2 hotspot waivers (R1 {40,41}, R1 {71,72}, R1 {72,73}, R2 {71-75}) dissolve. Each edit surfaces differentiation canon already wrote and matches the already-judged boss-roster row, so nothing here invents phase mechanics or antagonist form. Gate-A verified green.
• Canon support: [SRC: CVD §12.2 L654-664 the seven modes + the R1/R2 rule the edits satisfy] [SRC: CVD Pillar 3 L182 'flanking chapter intensification (Ch 37/39…)' — intensification is canon, no Velheim form assigned] [SRC: T4_HoV §8.2 L337 'Ch 37… Antagonist-family-presence canonical-absent' + RCM L4190 'no antagonist encounter at this chapter' — kills the old velheim_intrusion→direct mapping] [SRC: RCM L4375 Ch 40 'inscription-ability reading of corporate internal-communications and shipping-manifest archives… documented Disruption' = COMBAT L327 'Stealth encounters: infiltration and extraction missions'] [SRC: T22 L1170 'Ch 72… Dual-mechanic simultaneous-combat-and-vril-restoration' + VRIL §11.8 L360 'Ch 72 Realm War Phase 2 sacred-site substrate-restoration extension' + COMBAT L334-341 env-boss protocol = environmental] [SRC: boss_roster_rows.json Ch 37 attune/restraint, Ch 40 choose/expose+stealth-substrate, Ch 72 restraint-purge hybrid] [VERIFIED this session: Gate A PASS, exit 0, 65 R4 waivers intact, scratch candidate with exactly these edits]
• B — Approve CH_40 + CH_72; hold CH_37 as direct (status quo) — Take the two edits whose differentiation is written in a domain doc (COMBAT L327 for Ch 40; T22 L1170 + VRIL L360 for Ch 72) but keep Ch 37 direct pending a separate ruling.
• Canon support: [SRC: CVD Pillar 3 L182 flanking intensification could be read as combat at Ch 37] but [SRC: T4 L337 + RCM L4190 antagonist absence] make direct rest on a velheim_intrusion mapping canon denies. Rejected: leaving Ch 37 direct keeps the sharpest documented WGM-vs-CVD conflict open and the taxonomy [DECISION NEEDED] load-bearing; R2 windows across the crossing stay at the bare 3/5 minimum with zero headroom.
• C — Approve the three modes but also apply downstream intensity accents now (CH_64 valley / CH_73 peak) — Bundle the mode fixes with pacing-contour intensity changes that are tonally supported but not canon-stated.
• Canon support: Rejected on do-not-invent: [SRC: CVD §16.4] defers per-chapter intensity to the T2/spine layer; a CH_73 peak creates a 73-74-75 triple violating [SRC: HL_0048 'No three consecutive intensity-peaks'] (Ch 74 peak canon-anchored [SRC: CVD Pillar 2 L175]; Ch 75 HL_0100). Intensity accenting is Ruling-3 downstream fine-tuning, not this work order.

Recommendation — A — Approve all three edits + remove the 4 waivers; rule the 4 carried decisions per the recommendations below

How it fits the CVD. The package serves the CVD's mode-variety discipline directly. CVD §12.2 L664 is the constitutional rule the three edits satisfy: "Two consecutive chapters never share dominant combat type. Every five-chapter window contains three or more distinct combat-mode dominant chapters" — the four dissolved waivers were the only live violations of exactly this sentence, and after the edits every window across all three spans carries 4-5 distinct modes (verified). None of the three edits invents canon: each SURFACES differentiation the domain docs already state (CH_40 stealth = RCM L4375 corporate-comms infiltration mapped to COMBAT L327; CH_72 environmental = T22 L1170 dual combat-and-restoration + VRIL §11.8 L360 + COMBAT §9.2 protocol; CH_37 vril_management = the CVD Pillar 3 L182 flanking-intensification expressed as signature-hold-without-discharge, CVD §12.2 L660's vril_management definition verbatim, with the antagonist absence T4 L337/RCM L4190 require). The CH_37 edit RESOLVES a standing CVD-vs-WGM conflict: the prior "direct" rode velheim_intrusion→direct, but CVD L182 names the intensification without assigning Velheim a form and T4/RCM state antagonist absence — vril_management is the only classification consistent with CVD, T4, and RCM at once. The edits align with Pillar 8 (blended-mode bosses) and Pillar 3 (Hard Line combat routes through non-antagonist layers).

How it builds canon. Approving stands up the live topography layer as fully Gate-A-green net of only the 65 R4 (five-beats) waivers, which are spine-cascade authoring, not mode defects. It removes the last four mode-level waivers, converting them from "held for Josh" to canon-grounded resolved cells — so the Phase 6 spine cascade inherits a topography table with zero open mode alternatives across these five hotspots. It confirms the seven-slug CVD §12.2 taxonomy as the closed dominant_combat_mode enum that every future chapter classification depends on. And it locks three roster/topography alignments (the Ch 36-37-38 staircase, the Ch 59-64 counter-campaign rotation, the Realm-War per-phase structure) so roster and topography tell one story. The three genuinely Josh-exclusive items it surfaces (region order, Ch 62 op-typing, Ch 73 grade) are isolated and flagged, not silently resolved.

Why this option. A beats B because B's reason to hold CH_37 (weaker grounding) is backwards: CH_37 direct is the classification with a canon CONTRADICTION behind it (velheim_intrusion→direct against T4 L337 + RCM L4190 antagonist absence), while vril_management resolves the contradiction and matches the already-judged roster row (attune/restraint = CVD §12.2 L660 verbatim). Holding it keeps the span at zero R2 headroom and a load-bearing [DECISION NEEDED] open for no gain. A beats C because C bundles canon-stated mode fixes with canon-silent intensity accents that Ruling 3 explicitly sends downstream, and one of C's accents (CH_73 peak) would itself create an HL_0048 triple-peak violation — C spends hard-line headroom to add nothing witnessed. A is minimal-diff (three cells, the smallest set dissolving all four waivers), mechanically verified green, and every edit is canon-surfacing rather than canon-inventing — the exact discipline the cardinal rule demands.

Three-layer fit.

• Surface (L1): L1 SURFACE: the three edits make three flat spans playable. The Atlantic crossing becomes a designed three-beat staircase (Ch 36 perceive currents → Ch 37 hold-your-signature-as-density-swells → Ch 38 the seven-vril-seat Midpoint exam) instead of three quiet sails [SRC: boss_roster_rows.json Ch 36/37]. The modern arc opens with a kinesthetic anchor — Ch 40 becomes supply-chain infiltration gameplay, halving stealth's 11-chapter absence — instead of a five-chapter dialogue run (the audit's #1 range risk). The Realm War stops being three identical vril-combat beats and becomes hold-the-front / restore-the-sites-under-assault / bring-the-whole-kit-to-the-titan [SRC: boss_roster_rows.json Ch 71/72/73]. Failure stays graduated, never a fail-wall (a fumbled Ch 37 hold means the Convergence meets you 'fed' at Ch 38 [SRC: T1_Combat L346]).
• Mid (L2): L2 MID/DEPTH: each new mode makes an investigatable substrate playable. Ch 40's stealth layer IS the inscription-ability reading of corporate internal-communications and shipping-manifest archives [SRC: RCM L4375] — the documented worldview (conflict-minerals supply chain) becomes the encounter, not framing. Ch 72's environmental restoration surfaces the fairy-realm echo-source survey — sacred sites revealed as the sources of the human-world echoes encountered Ch 2-57 [SRC: T22 L1170] — so surface and depth share one chapter because canon put them there.
• Deep (L3): L3 DEEP: the deepest layer routes through the same mechanics untouched. Ch 72's environmental win condition is the §9.2 protocol (perception → vril-management → restraint → purgation) — the metaphysical-as-operational, never declared. No edit adds any Grand Sage hint or pre-Ch-76 surfacing; the show-don't-tell discipline (RCM L6683 'no pre-Ch 76 Grand Sage identity hints permitted') is preserved.

Test (gate). Apply exactly the three dominant_combat_mode cell edits (CH_37 direct→vril_management, CH_40 puzzle_routed→stealth, CH_72 direct→environmental) to the live Chapter_Index CSV, remove the four waivers (R1 {CH_40,CH_41}, R1 {CH_71,CH_72}, R1 {CH_72,CH_73}, R2 {CH_71,CH_72,CH_73,CH_74,CH_75}) from docs/topography_baseline.json, then run: python harness/check_chapter_topography.py _Expected:_ Gate A PASS, exit 0: 0 missing columns, 0 unpopulated, 0 bad values, 0 combat-absent, 0 unwaived rule violations, 0 stale waivers, exactly 65 R4 waivers remaining (79 nodes, Pro/Epi included). ALREADY VERIFIED this session on a scratch candidate with exactly these edits — result was PASS / exit 0 / 65 R4 waivers.

Adversarial review — HOLDS. I independently verified every load-bearing anchor by reading the cited canon (not trusting the recommendation's quotes) and reproduced the gate result on a scratch candidate. The recommendation survives adversarial review.

REFUTATION ATTEMPTS AND WHAT HELD:

1. INVENTED-CANON CHECK (cardinal rule) — HELD. Each of the three edits surfaces differentiation canon already wrote, not new phase mechanics:

• CH_72→environmental: VRIL [ACTIVE v2.2] L360 verbatim "Ch 72 Realm War Phase 2 sacred-site substrate-restoration extension"; T22 [ACTIVE v2.2] L1170 verbatim "Dual-mechanic simultaneous-combat-and-vril-restoration"; COMBAT §9.2 4-step protocol (perception→vril-mgmt→restraint→purgation) confirmed. Solid.
• CH_40→stealth: RCM L4375 verbatim "inscription-ability reading of corporate internal-communications and shipping-manifest archives"; COMBAT L327 "Stealth encounters: infiltration and extraction missions." Defensible (see stronger_option caveat).
• CH_37→vril_management: CVD §12.2 L660 verbatim "Vril-management combat (combat that routes through vril resource discipline)"; boss roster verb "attune" = vril-management. No invented mechanic.

2. THE CH_37 CANON-CONTRADICTION CLAIM — CONFIRMED. The old "direct" rested on velheim_intrusion→direct, but T4_HoV L337 verbatim "Ch 37 Final Approach. Antagonist-family-presence canonical-absent" and RCM L4190 verbatim "no antagonist encounter at this chapter per the no-encounter narrative discipline" both deny an antagonist. So "direct" (implying a Velheim combat opponent) is the classification with a genuine T4/RCM contradiction behind it; vril_management resolves it. The recommendation's "why A beats B" reasoning is correct, not backwards.

3. HIDDEN R1 ADJACENCY — CHECKED, HELD. I extracted the live mode sequence for both spans. CH_37's deeper reason for vril_management (beyond the contradiction) is that its boss-roster KIND is "environmental," and its neighbors CH_36 and CH_38 are both environmental — classifying Ch 37 environmental would create an env-env-env R3/R1 triple. vril_management is the only §12.2 slug that both fits the attune/restraint verbs AND avoids the collision. CH_40→stealth: CH_39=direct, CH_41=puzzle_routed, so no collision, {40,41} R1 dissolves. CH_72→environmental: CH_71 direct, CH_73 direct → env breaks both {71,72} and {72,73} R1 waivers and lifts the R2 {71-75} window from 2 to 3 distinct modes.

4. GATE CLAIM — INDEPENDENTLY REPRODUCED. I copied the live Chapter_Index to scratch, applied exactly the three cell edits, removed exactly the four waivers from a scratch baseline, and ran python harness/check_chapter_topography.py --candidate <scratch> --baseline <edited>. Result: Gate PASS, exit 0, 0 unpopulated, 0 bad values, 0 combat-absent, 0 unwaived violations, 0 stale waivers, exactly 65 R4 waivers. Matches the recommendation's test claim. (Note: current live ALSO passes with 69 waivers, so the real value of the edit is dissolving 4 held waivers, not fixing a red gate — the recommendation states this honestly.)

5. OPTION C KILL — CONFIRMED. CH_74=peak and CH_75=peak in the live table; HL_0048 (T0_Hard_Lines L49, CVD §17.10) verbatim "No three consecutive intensity-peaks." A Ch 73 peak makes 73-74-75 peak-peak-peak → hard-line violation. C is correctly rejected.

6. THREE-LAYER — HOLDS. Edits are pure combat-mode reclassifications touching zero narrative content; no pre-Ch-76 Grand Sage hint is introduced, so §17.9/show-don't-tell is untouched. L2 claims (Ch 40 archive-reading = investigatable substrate; Ch 72 echo-source survey = T22 L1170 human-world-echo-source recognition) are textually grounded.

7. AUTHORITY-HIERARCHY — CORRECT. The region-order conflict is REAL (verified both docs): WGM v2.3 gives Ch 59 Africa / 60 Asia / 61 Europe / 62 Americas / 63 Pacific; T4 §9.4 gives Ch 59 Oceania / 60 Europe / 61 Middle East / 62 Asia / 63 Americas. The recommendation flags this as Josh-exclusive and recommends WGM per-chapter contents (geography is WGM's domain under CVD>T1-domain), T4 op-type overlay retained — hierarchy-consistent and correctly NOT silently resolved. CH_62 low-risk-either-way confirmed (neighbors CH_61 stealth / CH_63 puzzle_routed leave R1/R2 legal whether Ch 62 is vril_management or environmental). HL_0104 (T0_Hard_Lines L105) confirms the Ch 40-41 op-type exception and the VRIL-standalone→environmental mapping the flag cites.

8. DISPOSITION — CORRECT, not too aggressive. await_josh_confirm is right: the three cell edits alone are Ruling-3-delegated grounded corrections, but they are bundled atomically with the region-order/CH_62/CH_73/taxonomy rulings which ARE Josh-exclusive. It does not apply_now on any invented canon. Safe to apply the instant Josh confirms.

Stronger option the adversary surfaced: Not a full alternative, but a sharper framing the recommendation under-states: at CH_40 the dominant-combat-MODE axis (§12.2, seven slugs) and the boss-KIND axis are orthogonal, and they point different directions. Ch 40's operational type is FINANCIAL/Moral-Choice (HL_0104: "FINANCIAL Sector 1 Boss approach Social"; RCM "Moral Choice boss type"; boss-roster verbs choose/expose, win=pacify) — none of which is stealth. stealth is defensible ONLY as the chapter's combat-texture emphasis (the archive-reading gameplay at RCM L4375), not as the boss kind. Textually, "infiltrate" as a boss verb actually appears at Ch 39 and Ch 44, not Ch 40. So CH_40→stealth is a sound, gate-clean judgment call, but it is a JUDGMENT call, not the uniquely-forced reading the "why" section implies. The recommendation should surface to Josh that Ch 40's stealth is a mode-emphasis choice sitting on a Moral-Choice boss (the two axes coexist), so he rules with eyes open. This does not change the verdict — the item is held for Josh, not applied — and stealth remains the strongest R1-dissolving choice for the {40,41} pair.

Corrections to fold in:

• Minor over-statement in the 'why' section: CH_40→stealth is presented as canon-forced, but Ch 40's boss KIND is Moral-Choice/FINANCIAL (HL_0104, RCM, roster verbs choose/expose) — stealth is the chapter combat-MODE emphasis on a Moral-Choice boss, an orthogonal axis. Recommend Josh be told the two axes coexist so the confirm is informed. Gate-clean and defensible; does not change the disposition.
• Framing precision on the gate 'value': the current live table ALREADY passes Gate A (with 69 waivers). The edits' value is dissolving 4 held waivers, not turning a red gate green. The recommendation does state this, but the surface phrasing 'stands up the live topography layer as Gate-A-green' could read as if the gate was previously failing — it was not.

Apply actions. On Josh's confirm, apply atomically: (1) registries/'T0_Chapter_Index [ACTIVE v1.2]'/'T0_Chapter_Index DRAFT v1.1.csv' — set dominant_combat_mode: CH_37 direct→vril_management, CH_40 puzzle_routed→stealth, CH_72 direct→environmental (three cells only; no intensity, no mode_variety_check changes). (2) docs/topography_baseline.json — remove the four waivers: R1 {CH_40,CH_41}, R1 {CH_71,CH_72}, R1 {CH_72,CH_73}, R2 {CH_71,CH_72,CH_73,CH_74,CH_75}; all 65 R4 waivers remain. (3) docs/proposals/topography_rows.json + docs/archive/TOPOGRAPHY_TABLE.md — mirror the three cells; rewrite the [DECISION NEEDED] flag text at CH_34/37/40/41/61/62/64/71/72/73 recording Josh's confirmations (CH_37 velheim_intrusion-mapping + intensification-expression resolved; CH_40/41 R1-waiver text deleted as superseded; CH_61/62/64 alternatives → confirmed; CH_71/72/73 hotspot-waiver → resolved), keeping the CH_41 Vimana-numbering and CH_73 intensity flags open; update the waiver-summary line. (4) If Josh rules region-order/CH_62/CH_73 differently from the recommendations, re-derive the CH_59-63 mode assignments (region order) or flip CH_62 to environmental (VRIL-standalone) BEFORE running the gate. (5) Re-run: python harness/check_chapter_topography.py — expect PASS, exit 0, 65 R4 waivers. (6) Commit recording Josh's rulings on the four carried decisions. Do NOT touch any boss_roster_rows.json row — no roster change is required by any selection.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] REGION ORDER (Josh-exclusive canon): T4_HoV §9.4 L401-409 (59 Oceania/Zealandia/Potlatch, 60 Europe, 61 Middle East, 62 Asia, 63 Americas, 64 Atlantis) vs WGM L747-786 (59 Africa, 60 Asia, 61 Europe, 62 Americas, 63 Pacific, 64 Underwater) — a genuine conflict, verified in both docs. Applied table + roster follow WGM regions with T4 op-types. RECOMMENDATION: rule for WGM's per-chapter CONTENTS (the vril-site recall sits in WGM's Ch 62 Americas entry; Zealandia in WGM's Ch 63), because the mode rotation and roster both ride WGM contents and WGM is the geography domain authority; the op-type overlay stays T4's. If Josh rules for T4's order, the 59-63 modes re-derive with the moved contents — rotation SHAPE survives, chapter SLOTS shift. This is the one item that must be ruled to lock CH_59-63.
• [DECISION NEEDED - JOSH] CH_62 OP-TYPING (Josh-exclusive antagonist canon): POLITICAL ACCESS (applied vril_management stands, grounded in WGM L771-777 vril-gathering-site recall) vs VRIL-standalone (HL_0104 maps it to 'Environmental combat with sacred-site restoration' → mode flips to environmental). RECOMMENDATION: POLITICAL ACCESS / keep vril_management — it stands on the chapter's own stated site-recall content and T4 §9.4 frames Ch 59-65 as the Political Access counter-campaign primary manifestation; R1/R2 stay legal either way, so low-risk whichever Josh picks.
• [DECISION NEEDED - JOSH] CH_73 INTENSITY GRADE (Josh-exclusive pacing ruling): T22 L1172 'Game's largest single combat encounter' argues for peak vs the standard grade forced by HL_0048. RECOMMENDATION: keep STANDARD — scale is encounter design, grade is pacing contour; a Ch 73 peak creates a 73-74-75 triple violating HL_0048 (Ch 74 peak canon-anchored CVD Pillar 2 L175; Ch 75 HL_0100). Grade-only; the mode edit CH_72→environmental is independent of it.
• [DECISION NEEDED - JOSH] TAXONOMY SLUG CONFIRM (carried from 0.D.1): the seven CVD §12.2 slugs (direct, stealth, puzzle_routed, environmental, vril_management, familiar, astral) as the closed dominant_combat_mode enum, and their lowercase spellings. RECOMMENDATION: CONFIRM — the seven are verbatim CVD §12.2 L656-662, the only list CVD attaches to the R1/R2 rule; all three edits stay inside the set. Spellings are a representational choice (canon gives phrases); flag if Josh wants different slugs.
• [DECISION NEEDED - JOSH] DISPOSITION NOTE: the three cell edits alone are Ruling-3-delegated grounded corrections and gate-verified green, but the package is presented as one atomic confirm tied to the region-order/CH_62/CH_73 rulings (which are Josh-exclusive), so the whole is await_josh_confirm rather than apply_now — safe to apply the instant Josh confirms.

---

6. Confirm WGM combat-layer wording (Ch 38/52/57) + VS_CH57_001 sub-decisions

• Key: WGM_COMBAT_LAYER_CH38_52_57 Type: creative Disposition: await_josh_confirm

Options

• A — Confirm all four WGM edits + the two residues as written; split the VS_CH57_001 row and the L371 contradiction out as separate rulings — Apply the four byte-precise WGM string swaps (L528/674/725/729) plus the T1_Vril §11.1 L314 and T1_Languages L66/L233 residues, exactly as the proposal drafts them — the environmental/atmospheric-boss (§9.2 protocol) framing for Ch 38/52/57, with all Ch 57 combat routed through non-person modes (environmental/astral/vril-substrate-defense). These are grounded reconciliations of a stale 'No combat' migration wording against Ruling 1 + the live hard-line rows. Hold the VS_CH57_001 registry row (site_class + canonical_status) and the T1_Combat L371 fauna/training contradiction as their own Josh sub-rulings, because those are canon-taxonomy invention and a hard-line-adjacent conflict respectively — not the same disposition class as the wording swaps.
• Canon support: [SRC: CVD Pillar 3 L181 # "Previous 'no combat' framing is rejected as creatively fatal... the combat layer routes through Velheim intrusion or astral/dream-world threats inside vril-sanctified spaces"] [SRC: CVD §16.2 L877 # "Hard Line chapters... protect cultural sanctuary through Velheim intrusion combat or astral and dream-world combat — never through combat absence. Combat absence is creatively fatal"] [SRC: CVD §17.1 L933 # Ch 57 "no combat against community representatives. Combat arrives through environmental and astral mode"] [SRC: HARDLINES HL_0050/HL_0051/HL_0106/HL_0008 # relief-tier combat canonical; Ch 57 environmental/astral/vril-substrate-defense; no antagonist presence] [SRC: T1_Combat §Environmental-and-Atmospheric-Boss-Mechanic L332-358 # 4-step protocol, cannot be defeated through damage, Ch 38/52/57 instances] [SRC: T1_Vril §11.1 L314 + T1_Languages L66/L233 # the §17.7 mis-citations that are the same defect class]. Verified: all three WGM replacements route Ch 57 combat only through non-person modes, honoring §17.1 L923 'Never combat against community representatives.'
• B — Confirm only the three WGM combat-layer edits (L528/674/725) now; defer the L729 cultural-discipline edit, both residues, and the whole registry row — Narrow scope to the three Type-line swaps that most directly execute Ruling 1, and hold everything that touches the cultural-discipline bullet (L729), the cross-doc residues, and the new registry instantiation for a later batch. Rationale would be minimizing surface area on the highest-care chapter. In practice this is strictly worse: it leaves L729 saying flatly 'No combat, no antagonist family — these are absolute' one line below the reframed L725, creating a fresh internal contradiction inside the same chapter entry, and it leaves the §17.7 mis-citations (a known defect class Josh already saw corrected for Integrity_Paths) unfixed.
• Canon support: [SRC: WGM Ch 57 L725 vs L729 # the two lines both currently assert blanket 'No combat'; reframing one without the other is self-contradiction, the exact defect GF-1 corrected in T3_Creatures] [SRC: CORPUS_SWEEP GF-3/§2.3 # the §17.5/§17.7 mis-citation class was already ruled a grounded fix elsewhere] — this option's own grounding argues against it.
• C — Rewrite the replacement wording to be terser (drop the inline [SRC]-style clause pileup) before confirming — Keep the same canonical substance but tighten the NEW strings — several run long with stacked citations to HL_0106/§9.2/four-step-protocol inside the prose bullet, which is heavier than the surrounding WGM entries' register. A terser form (e.g. Ch 38: 'Type: Transformation. Combat layer per Pillar 3 — the triple-plate-junction vril convergence as a Tier A environmental boss resolved through the four-step protocol, never through damage; no antagonist family presence per HL_0106.') carries identical canon with less in-line machinery.
• Canon support: [SRC: CVD §17.14 / §20 # paste-ready formatting: plain prose, affirmative framing — the register the WGM entries otherwise hold] [SRC: WGM Ch 38 L525-533 # surrounding bullets are single-clause; the drafted NEW string is materially denser than its neighbors]. No canon opposes the substance; this is a register/paste-quality refinement, not a canon change.

Recommendation — A — Confirm all four WGM edits + the two residues as written; split the VS_CH57_001 row and the L371 contradiction out as separate rulings

How it fits the CVD. The reconciliation is the direct instrument of Josh's Ruling 1 (L1 action-packed everywhere; the hard line wins over WGM 'No combat'; the environmental-boss protocol is the playable layer for protected/relief chapters). Every replacement string traces to canon that already exists and outranks the stale WGM wording: Pillar 3 [SRC: CVD L181] and §16.2 [SRC: CVD L877] both name 'combat absence is creatively fatal' as a pillar violation, and §17.1 [SRC: CVD L933] fixes the exact modes Ch 57 uses (environmental + astral). The 'No combat' strings on WGM L528/674/725/729 are pre-Ruling-1 migration residue that contradicts CVD directly — under §17 'where two canon instructions appear to conflict, the hard line wins' [SRC: CVD L913], so WGM must yield to CVD. Critically for the highest-care chapter: the drafted Ch 57 wording routes combat ONLY through non-person modes (environmental/astral/vril-substrate-defense) and explicitly retains 'No combat against community representatives / No antagonist family presence' — it honors §17.1's inviolate-representatives line [SRC: CVD L923] while removing only the blanket combat-absence claim that §17.1 itself already contradicts two clauses later.

How it builds canon. It converts four contradiction sites into cascade-ready anchors: the Phase 6 spine's Grid 8 boss-variety rubric and the L1-playable-boss ruling both consume the WGM Type line, and today that line tells a spine author 'No combat' at a Tier-A boss chapter (Ch 38) [SRC: T1_Combat L213 # Ch 38 hard-line-spine boss anchor]. Fixing it unblocks the boss-roster proposal work order for Ch 38/52/57 and removes the §17.7 mis-citations that would otherwise mis-route any consumer reading Ch 38 combat status through T1_Vril or T1_Languages. The VS_CH57_001 row (once Josh rules its two sub-fields) instantiates the first structured Ch 57 vril-site anchor grounded in the §5.7 recharge-class + §12.5 songline-anchor canon, feeding the T0_Vril_Site_Registry two-column spine scatter. It builds forward without inventing world facts — every claim is a restatement of CVD/T1 canon already on the page.

Why this option. A beats B because B manufactures a NEW intra-chapter contradiction: reframing L725 while leaving L729's 'No combat... these are absolute' one line below is precisely the self-contradiction defect the corpus sweep flagged and corrected elsewhere (GF-1) [SRC: CORPUS_SWEEP GF-1 L91-100]. The four WGM edits plus the two residues are one coherent reconciliation unit — the §17.7 mis-citations in T1_Vril L314 / T1_Languages L66/L233 are the same already-ruled defect class as the applied Integrity_Paths §17.5 fixes, so leaving them is arbitrary. A beats C because C's terser rewrite, while a legitimate register improvement, changes bytes on the single most culturally-sensitive chapter's entry for a paste-quality gain, not a canon gain — and the density is defensible: a Tier-A hard-line boss line earns its citations. If Josh wants the terser register, that is a trivial follow-up edit, not a reason to hold the reconciliation. A's discipline of splitting the registry row and the L371 fauna contradiction into their own rulings is what keeps A safe: it does not smuggle canon-taxonomy invention (site_class) or a hard-line-adjacent conflict resolution into a wording confirmation.

Three-layer fit.

• Surface (L1): L1 SURFACE: the fix is exactly what makes Ch 38/52/57 playable rather than cutscene-passive — the environmental/atmospheric boss (perceive → manage vril → restrain → purge, never damage) is the operational combat layer at chapters that currently read 'No combat.' Ch 38's triple-plate-junction convergence and Ch 57's Dreaming-site protocol become fightable content; the action-adventure arc stays complete and satisfying on the surface [SRC: T1_Combat L332-358].
• Mid (L2): L2 MID/DEPTH: the reframing preserves the investigatable substrate — Ch 57's site is a §5.7 recharge source that does not respond to negative-vril signatures at all, and the elder reads vril signature directly [SRC: T1_Vril §5.7 L186, §1 L25]; the journal-public beat and songline-as-living-inscription (referenced, never reproduced) remain the depth-player's evidence trail. Nothing in the edit closes off the documented-worldview layer; it opens the combat layer that the depth player also reads as vril-substrate defense.
• Deep (L3): L3 DEEP: the cultural-protection absolutes are the air the chapter breathes — the edit strengthens, not weakens, the L3 truth by keeping 'no combat against community representatives / no antagonist family presence' as inviolate while removing the blanket claim that would have flattened Ch 57 into non-interactive reverence. The convergence-thesis (Pattern across disconnected cultures) is served, never declared: Ch 57 stays the single chapter where a human names the child accurately, untouched [SRC: WGM Ch 57 L728].

Test (consistency). The stated TEST ('scan gate green') is necessary but NOT sufficient and must not be relied on alone: the migration scanner only checks lowercase_paragraph + consecutive_header and is blind to 'No combat' wording, so the WGM swaps are byte-neutral to it and it will stay green whether or not the edits are correct (verified: harness/scan_migration_integrity.py scans only those two categories; 'No combat' appears nowhere in it). The real test is (1) run python harness/run_gates.py and confirm all five gates (fidelity/scan/topography/lint/anchors) still PASS after the edits — fidelity in particular, because the new VS_CH57_001 row must be declared in docs/fidelity_baseline.json added_rows or fidelity FAILs; and (2) consistency-anchor re-verification: after the swaps, grep the three docs (WGM, T1_Vril, T1_Languages) for any surviving 'No combat' that is unscoped (blanket, not 'against community representatives'), and confirm zero Ch 57 replacement clause names a person-directed mode. These I performed against the current text; they must re-pass post-edit. _Expected:_ run_gates.py exits 0 (all five PASS) with the VS_CH57_001 row declared in fidelity_baseline added_rows; zero unscoped 'No combat' strings remain in WGM/T1_Vril/T1_Languages for Ch 38/52/57; every Ch 57 combat mode in the replacement text is non-person (environmental/astral/vril-substrate-defense).

Adversarial review — HOLDS. REFUTATION ATTEMPT — every load-bearing anchor independently read; the recommendation held.

WHAT I VERIFIED (all confirmed accurate against disk):

• WGM byte targets L528 ("Type: Transformation. No combat."), L674 ("Type: Meditative. No combat — game's quietest beat."), L725 ("...Most protected scene... No combat. No antagonist family presence. Ch 57 hard line per CVD §17.1..."), L729 ("...No combat, no antagonist family — these are absolute.") — all match the proposal exactly. The L725/L729 self-contradiction that Option B would create is real (both lines assert blanket "No combat" one entry apart), so A correctly beats B.
• CVD Pillar 3 L181 ("Previous 'no combat' framing is rejected as creatively fatal... routes through Velheim intrusion or astral/dream-world"), §16.2 L877 ("never through combat absence. Combat absence is creatively fatal"), §16.4 L905 (Ch 57 "Combat arrives through environmental and astral mode"), §17.1 L923 ("Never combat against community representatives... inviolate") and L933 (Ch 57 "Combat arrives through environmental and astral mode"), and the "hard line wins" clause L913 — all quoted verbatim-accurate. The authority hierarchy holds: WGM's stale "No combat" yields to CVD, correctly.
• T1_Combat §9.2 L332-358 environmental-boss 4-step protocol + Ch 38/52/57 instances — confirmed, grounding the L1-playable framing.
• Residues: T1_Vril §11.1 L314 and T1_Languages L66/L233 do say "No combat per CVD §17.7". Verified §17.7 (L999-1014) is PURELY naming hard lines — it does NOT govern combat presence. So the §17.7→Pillar-3 combat re-citation is a grounded fix of a genuine mis-citation, same defect class as the ruled Integrity_Paths §17.5 fixes. (Caveat below.)
• Registry: T0_Vril_Site_Registry Sheet1.csv uses only LOCKED / CASCADE_CONFIRMED — "PROPOSED" is genuinely out-of-vocabulary; and site_class has no songline/distributed-path value. Both sub-flags are real open canon-taxonomy decisions, correctly held for Josh, not smuggled into the wording confirmation. No VS_CH57 row exists yet (count 0), so the row is a clean add.
• HARD-LINE ROWS decisively corroborate: HL_0008 gives an EXHAUSTIVE Ch 57 mode list "environmental hazards plus astral mode plus vril-substrate-defense"; HL_0051 names Ch 57 the canonical exception routing "environmental, astral, and vril-substrate-defense"; HL_0106 confirms Ch 38/52/57 no antagonist presence; HL_0077 lists Ch 38 as a Tier-A Solo boss anchor (so WGM "No combat" at Ch 38 contradicts a canonical boss designation); HL_0080 lists Ch 38/57 as §9.2 environmental-boss instances. The reconciliation is not invention — it is a restatement of canon already locked in the hard-line registry.

ATTACKS THAT FAILED:

1. Invented canon? No. Every replacement clause traces to CVD/hard-line rows already on the page. The two items that WOULD be invention (site_class taxonomy, canonical_status value) are explicitly deferred, not asserted.

2. Authority-hierarchy error? No. The proposal makes the lower doc (WGM) yield to CVD, which is correct per L913. It does not treat any T99/draft as binding.

3. §17 violation smoothed over? No — the opposite. The Ch 57 wording RETAINS "no combat against community representatives / no antagonist family presence" (honoring §17.1 L923 inviolate-representatives) and removes only the blanket combat-absence claim that §17.1 L933 itself already contradicts.

4. Three-layer claims: L1 (environmental boss makes Ch 38/57 playable), L2 (recharge-site substrate preserved), L3 (protection absolutes strengthened, convergence never declared) all hold against the read text. The "single chapter where a human names the child accurately" (WGM L728) is genuinely untouched.

5. Disposition too aggressive? No — await_josh_confirm is correct: the edits land in _source (frozen ground truth) and instantiate a registry row with two open sub-fields, exceeding apply_now's gate-safe bar even though Ruling 1 delegates the reframing CLASS. The scan-gate-only TEST is correctly flagged as necessary-but-insufficient (the scanner is blind to "No combat" wording); run_gates.py + the fidelity_baseline added_rows declaration is the real gate.

The recommendation is canon-faithful, anchors check out, and no material flaw changes the verdict. Recommend A, disposition unchanged at await_josh_confirm.

Stronger option the adversary surfaced: none — A is the right disposition. The only improvement is internal to A: (1) foreground HL_0008's exhaustive-mode-list and HL_0077's Tier-A-boss-at-Ch-38 rows as the primary authority (stronger and more direct than the prose citations the proposal leads with), and (2) upgrade the L371 sub-flag from a neutral defer to a canon-leaning recommendation per the corrections above. Neither is a different option; both strengthen A.

Corrections to fold in:

• Residue-fix precision (T1_Vril L314, T1_Languages L66/L233): the proposal calls these a '§17.7 mis-citation.' Accurate for the COMBAT clause, but §17.7 L1006 ('Ch 38: Azores. Canonical Atlantis position') legitimately anchors the Azores naming/position those same lines reference — and the registry itself uses CVD §17.7 as VS_CH38_001's source_authority. The apply-step must re-cite §17.7→Pillar-3 ONLY on the 'No combat' clause and preserve any §17.7 naming/position reference in the same sentence, or it will strip a valid citation. This is an apply-precision note, not a flaw in the recommendation.
• L371 sub-flag is under-recommended, not mis-handled: the proposal frames the T1_Combat L371 fauna/training contradiction as a two-sided Josh coin-flip ('which side is canonical'). The canon tilts decisively toward RCM's 'no fauna combat': (a) §4.5 L151 — the very section L371 cites as authority — names only Velheim-intrusion/astral for Hard Line chapters and does NOT contain the fauna/training modes L371 attributes to it (a misattribution); (b) HL_0008 gives an EXHAUSTIVE Ch 57 mode list (environmental/astral/vril-substrate-defense) with no person-directed fauna combat; (c) HL_0072 documents fauna at Ch 57 as a Layer-2/3 vril-signature BEHAVIOR substrate, not a combat mode. The proposal should carry a leaning recommendation (fix L371 toward RCM, add 'no fauna combat' to the WGM L729 replacement) rather than presenting it neutral. Disposition (defer to Josh) is still correct because it is hard-line-adjacent; only the framing should be sharpened.

Apply actions. On Josh's confirm (the proposal is authored NOT-applied by design, and edits land in _source which is frozen ground truth — so this is await_confirm even though Ruling 1 delegates the reframing class; the byte-edits to canonical _source T1 docs plus a registry-row instantiation with two open sub-fields exceed apply_now's gate-safe/data-model bar): 1) Apply the four WGM exact-string swaps at L528/674/725/729 via exact-match tooling, preserving the two trailing spaces on L528/674/725 and the no-trailing-space + 'Ch 57\\.' escape on L729 (verified byte-exact against disk). 2) Apply the two residue swaps: T1_Vril §11.1 L314 and T1_Languages L66 + L233, replacing the 'No combat per CVD §17.7' framing with the Ch 38 environmental-boss wording and correcting the §17.7→Pillar-3 mis-citation. 3) Add the VS_CH57_001 row to registries/T0_Vril_Site_Registry [ACTIVE v0.1]/Sheet1.csv ONLY after Josh rules the two sub-fields (see open_subflags): resolve canonical_status to LOCKED or CASCADE_CONFIRMED (PROPOSED is not in the registry's two-value vocabulary) and resolve site_class (proposed vortex_anchor on the Gebel Barkal/Delphi natural-form precedent, or a new class e.g. songline_path — note the Schema Dictionary already canonizes 'songline_living_inscription' as a writing_system_type enum at L339, a precedent for a distinct songline taxonomy); then declare the row in docs/fidelity_baseline.json added_rows in the same commit. 4) Route the T1_Combat L371 fauna/training-combat contradiction to Josh: if he confirms RCM's 'no fauna combat' (CVD §17.1 + HL_0051 list only environmental/astral/vril-substrate-defense), edit L371 to drop the person-to-animal and training-combat modes and optionally add 'no fauna combat' to the WGM L729 replacement; if he rules fauna/training canonical, fix RCM L5465 instead. 5) Run the standing normalization gate (session edit + fresh-context adversarial critic) before commit, then python harness/run_gates.py must exit 0. 6) Commit with the WGM/Vril/Languages edits + the declared registry row together.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] VS_CH57_001 site_class — canon-taxonomy decision: reuse vortex_anchor (natural-form vril-concentration precedent: Gebel Barkal, Delphi) OR mint a new class (e.g. songline_path) since a songline is a distributed living-inscription path, not a point site. Extending the registry's site_class vocabulary is canon-taxonomy invention → Josh's call. Note: site_class is NOT enumerated in the Schema Dictionary (de-facto vocabulary from existing rows), but Schema Dictionary L339 already canonizes 'songline_living_inscription' as a writing_system_type enum, a precedent for distinct songline taxonomy.
• [DECISION NEEDED - JOSH] VS_CH57_001 canonical_status — the registry vocabulary is only LOCKED / CASCADE_CONFIRMED; 'PROPOSED' in the draft is a placeholder not in-vocabulary. Josh must pick LOCKED (grounded in a CVD §17.1 hard line + T1_Vril §5.7/§12.5) or CASCADE_CONFIRMED per his registry-status convention.
• [DECISION NEEDED - JOSH] T1_Combat L371 fauna/training-combat vs CVD §17.1 + HL_0051 + RCM L5465 — a real hard-line-adjacent contradiction: L371 lists 'person-to-animal with documented dangerous fauna' and 'training combat with documented martial traditions' as Ch 57 modes, but CVD §17.1 L933 + HL_0051 name only environmental/astral(/vril-substrate-defense), and RCM L5465 states 'no fauna combat' as a §17.1 absolute. Josh rules which side is canonical; the fix location and whether WGM L729 gains 'no fauna combat' both depend on his ruling.
• [DECISION NEEDED - JOSH] thread_anchors for VS_CH57_001 — draft picks T01|T05|T17; T02 (Rainbow Serpent non-physical-completion) and T21 (music/sound cosmology) are defensible additions. Thread assignment is chapter/thread canon → Josh's call if he wants the row's anchor set widened.

---

7. Pacing-philosophy rulings: R4 all-five-beats binds (spine authors gaps); Epilogue is R6/Ruling-1-exempt (non-playable bookend); Pro/Epi split — Prologue IN, Epilogue OUT of R1/R2/R3 windows

• Key: PACING_PHILOSOPHY_R4_EPI_PROEPI Type: creative Disposition: await_josh_confirm

Options

• A — R4 all-five binds; spine authors gaps. Epilogue R6-exempt (non-playable, combat=N/A). Prologue IN windows, Epilogue OUT (asymmetric). — Strict R4: CVD §4.5/§16.2/Pillar-11 verbatim 'every chapter contains combat+puzzle+learning+character+transformation beats' binds as ALL FIVE; the 65 current waivers are legitimate substrate-gap placeholders the Phase 6 spine dissolves chapter-by-chapter as it authors the missing beats. Epilogue is exempted from the never-waivable R6 combat gate and from Ruling-1's combat=Y default because canon makes it a path-invariant, non-playable unseen-reader journal frame (fade to black) — combat is set N/A, not Y. Prologue is a PLAYABLE tutorial node so it stays inside R1/R2/R3 adjacency windows; the Epilogue, being non-playable and post-finale, drops OUT of the windows.
• Canon support: [SRC: CVD §4.5 L151] 'Every chapter contains combat, puzzle, learning, character, and transformation beats proportional to chapter length' — verbatim, five named beats. [SRC: CVD §16.2 L875] restates identically. [SRC: CVD Pillar 11 L237] restates identically ('proportional to chapter length'). Thrice-stated → strict reading is the plain-text reading. [SRC: CVD §16.2 L875] 'proportional to chapter length' + 'Long chapters earn their runtime through variety' = the SCALE varies, not the SET — a short chapter carries all five lightly, it does not drop beats. [SRC: RCM L6968] Epilogue = 'unseen-reader frame: the healer's child's journal open on a table, an unseen reader turning the pages, fade to black … Path-variant Epilogue content is not authored' → non-playable, path-invariant, no combat surface exists. [SRC: HARDLINES HL_0045] 'the unseen-reader epilogue play identically across all three endings … Words are understood, not spoken' → confirms path-invariant bookend, not an encounter. [SRC: RCM L6937] Prologue = 'the playable figure at the Prologue is the Elven King at full sovereign capability — playable here' → Prologue IS playable (unlike Epilogue). [SRC: CVD §16.1 L871] 'Three consecutive intensity-peaks are not permitted' — an adjacency rule that only means something for playable, sequenced chapters.
• B — Weak R4 (no-single-mode-dominance) binds. Epilogue combat=Y stands (Ruling-1 default). Pro+Epi both IN windows (status quo default). — Read R4 as the weaker Pillar-11 executable-test standard: a chapter fails only on 'single-mode dominance', so a chapter needs mode VARIETY (no coasting) but not literally all five beats. This retires the 65 waivers immediately (most chapters already clear no-single-mode). Keep Epilogue combat=Y as the Ruling-1 'combat everywhere' default, and keep both bookends inside the pacing windows (the current harness default).
• Canon support: [SRC: CVD Pillar 11 L239] executable test: 'Chapter scoring single-mode dominance in Mode Variety Check = pillar violation' — the ENFORCEABLE test is single-mode dominance, not all-five. [SRC: CVD §12.2 L664] the 'two consecutive / five-window' machinery is defined over the SEVEN COMBAT modes, a different axis than the five narrative beats — arguable the five-beat line is aspirational framing, not a per-chapter gate. [SRC: CANON_AUDIT / Known-issue-4 Ruling 1] Josh ruled 'L1 must be action-packed/playable in every chapter' → combat=Y everywhere is the safe default, incl. Epilogue. AGAINST: [SRC: CVD §4.5 L151] verbatim text says 'contains … beats', not 'avoids single-mode dominance'; the weak reading rewrites the canon sentence.
• C — R4 all-five binds AND all-five is a hard gate NOW (no waivers). Epilogue exempt. Pro/Epi both OUT of windows. — Strict R4 with no waiver mechanism — every chapter must show all five beats immediately or the gate is red. Epilogue exempted as in A. Both Prologue and Epilogue excluded from the adjacency windows as non-'Ch 2-77' nodes.
• Canon support: [SRC: CVD §4.5/§16.2/Pillar-11] same verbatim basis as A. AGAINST (why not-now): [SRC: LEDGER_BATCH1 L60-62] the 65 waivers exist because mode_variety_check is [POPULATE→T2] substrate the spine has not authored yet — turning R4 red now would block the whole cascade on data that Phase 6 exists to produce. [SRC: RCM L6937] Prologue is PLAYABLE and sequenced (Prologue→Ch1→Ch2) → excluding it from windows discards real adjacency canon (§16.1 topography 'across the full arc').

Recommendation — Coherent-package: R4 all-five binds + spine authors the missing beats (65 waivers shrink through Phase 6); Epilogue is R6/Ruling-1-EXEMPT (combat set to N/A, not Y); Prologue stays IN the pacing windows, Epilogue drops OUT (asymmetric --exclude flag)

How it fits the CVD. Option A honors the CVD's own authority hierarchy and its verbatim text. R4: the five-beat sentence is stated THREE times verbatim [SRC: CVD §4.5 L151; §16.2 L875; Pillar 11 L237] and every instance carries 'proportional to chapter length' — the CVD's own scaling clause, which resolves the short-chapter objection: scale the beats down, never drop them [SRC: CVD §16.2 L875 'Long chapters earn their runtime through variety']. Reading it as mere no-single-mode-dominance (Option B) rewrites a thrice-locked canon sentence, which the do-not-invent rule and CVD-wins-conflicts hierarchy forbid. The spine-authors-the-gaps mechanic fits §16's design intent that pacing 'serves Layer 1 surface satisfaction across the 30-million-word arc' [SRC: CVD §16 L859] — the waivers are honest 'not-yet-authored' markers, not exceptions. Epilogue exemption fits because the CVD's combat hard lines (HL_0050/0051, Ruling 1) are about CHAPTERS with a playable surface; the Epilogue is canonically a non-playable, path-invariant bookend [SRC: RCM L6968; HARDLINES HL_0045] — forcing combat=Y there manufactures a fictional encounter the CVD explicitly says is 'not authored'. The Prologue-IN/Epilogue-OUT asymmetry tracks the one canonical difference between the bookends: the Prologue is 'playable here and only here' [SRC: RCM L6937], the Epilogue is fade-to-black — so §16.1 adjacency topography ('three consecutive peaks') applies to the Prologue's sequenced position but is meaningless for a post-finale non-playable frame.

How it builds canon. Stands up three durable spine invariants Phase 6 consumes directly. (1) R4-all-five becomes the mode_variety_check acceptance test each /spine-chapter run must satisfy or waive-with-reason; the 65 waivers become a burn-down ledger whose count is a real progress metric (waivers→0 == mode-variety coverage complete). (2) A canonical NODE-CLASS distinction is instantiated: playable-sequenced nodes (Prologue + Ch 2-77) vs non-playable-bookend nodes (Epilogue) — this is reusable data-model canon (the Epilogue also legitimately carries combat=N/A, boss=none, no dominant_combat_mode, matching RCM 'no corpus, game-canon narrative-frame'). (3) The harness gate contract is pinned: R6 (never-waivable combat-present) applies to playable nodes only; R4 is waivable-with-reason; R1/R2/R3 windows run over Prologue..Ch77. This removes the last three [DECISION NEEDED] cells blocking the topography table from going fully green as the cascade's pacing spine.

Why this option. A beats B because B's 'no-single-mode-dominance' reading contradicts the verbatim CVD sentence stated three times — that is silent canon-narrowing, exactly what the cardinal rule forbids; and B keeps a semantically-false Epilogue combat=Y that canon [RCM L6968] says cannot exist. A beats C because C makes R4 a hard red gate NOW, which would block the cascade on mode_variety_check data that is still [POPULATE→T2] substrate the spine itself must author [SRC: LEDGER_BATCH1 L60] — the waiver mechanism is the correct instrument for 'true canon target, not-yet-populated', letting the gate stay green-with-waivers while the burn-down happens inside Phase 6. C also wrongly drops the PLAYABLE Prologue from adjacency windows, discarding real §16.1 topography. A is the only option that (i) preserves the verbatim five-beat canon, (ii) keeps the cascade unblocked, (iii) refuses to fabricate Epilogue combat, and (iv) draws the Pro/Epi line on the one attribute canon actually distinguishes them by (playable vs not).

Three-layer fit.

• Surface (L1): L1 gains the strongest guarantee: R4-all-five means every playable chapter delivers combat AND puzzle AND learning AND character AND transformation beats — no 'talking chapters', no coasting [SRC: CVD §4.5 L151 'The game does not have talking chapters or puzzle chapters']. The Epilogue exemption protects L1 too: the surface bookend stays the intended quiet fade-to-black, not a jarring forced fight after the finale. Prologue-IN keeps the playable tutorial inside the mountain-range pacing the player actually experiences.
• Mid (L2): L2 is unaffected as infrastructure but benefits: the five-beat discipline guarantees each chapter carries a learning beat, which is the hook for the investigatable substrate (journal/Grimoire/inscription) L2 players pursue [SRC: CVD §16.2, learning beat]. The Epilogue's non-combat frame is exactly the L2 payoff surface — the journal 'open on a table' is the accumulated Layer-2 artifact being read back [SRC: RCM L6968].
• Deep (L3): L3 is untouched and protected. The Epilogue's path-invariance ('play identically across all three endings, words understood not spoken') is a Layer-3 show-don't-tell instrument [SRC: HARDLINES HL_0045; CVD §17.9] — forcing combat onto it would risk declaring/dramatizing what must stay unspoken. Keeping it a non-playable frame preserves the single-canonical-surfacing discipline.

Test (gate). After applying: (1) set Epilogue mode_variety_check combat -> N/A handling by adding Epilogue to an R6-exempt bookend set and setting its dominant_combat_mode/intensity as non-playable; (2) run the topography gate twice to confirm the asymmetry and that R4 waivers are the only remaining R4 signal. Commands: python harness/check_chapter_topography.py <chapter_index_csv> (default include-pro-epi) then python harness/check_chapter_topography.py <chapter_index_csv> --exclude-pro-epi. Also run the full gate roster python harness/run_gates.py and the stress test python harness/stress_test_registry.py (R4 'four-of-five fires' and R6 cases must still pass). _Expected:_ Prologue participates in R1/R2/R3 in the default run and is dropped only under --exclude-pro-epi; Epilogue never triggers R6 (combat-absent) because it is bookend-exempt, and never contributes a spurious R1/R2/R3 adjacency; R4 reports exactly the 65 waived beat-gaps (waivable, not red); Gate A stays green-with-waivers; stress test R4/R6 cases green. No new bad/unpopulated values introduced.

Adversarial review — FLAGGED FOR REVISION. Independently read every load-bearing anchor. R4 half holds; the Epilogue-exemption half does not, and it is decisive.

HELD (verified): The five-beat sentence is stated three times verbatim, each with "proportional to chapter length" [SRC: CVD §4.5 L151; §16.2 L875; Pillar 11 L237] — R4-all-five is the plain reading; Option B's "no-single-mode-dominance" rewrites a thrice-locked sentence. The harness already encodes R4 as ALL-FIVE, waivable per chapter [SRC: check_chapter_topography.py docstring L35-46], with the 65-waiver burn-down matching [SRC: LEDGER_BATCH1 L60-62; PHASE0_LESSONS L31-32]. Prologue-IN is grounded: Prologue is playable "here and only here" [SRC: RCM Prologue block; HL_0099]. parse_variety accepts only Y|N — subflag accurate. Disposition await_josh_confirm is right.

BROKE (Epilogue combat=N/A, R6-exempt, OUT of windows):

1) OVER-READ AS GROUNDED CANON. The recommendation claims canon "makes the Epilogue non-playable / combat not authored." Canon actually says PATH-VARIANT content is not authored: [SRC: HL_0045] "Path-variant epilogue content is not authored"; [SRC: RCM Epilogue block] "Path-variant Epilogue content is not authored per CVD §17.9". That is path-INVARIANCE (one shared epilogue across three endings), not non-playability, not combat-absence. The leap from "path-variant not authored" to "non-playable, combat=N/A" is unsupported inference dressed as a [SRC] anchor.

2) HARD-LINE COLLISION. The Epilogue is one of the 18 affirmed hard-line nodes [SRC: CASCADE_RUNBOOK L18 "Prologue, ... Ch 74, 76, 77, Epilogue"]. Per [SRC: HL_0051] a hard-line chapter "without combat layer at all is a pillar violation. Combat absence is creatively fatal ... via combat, not by combat absence." Setting Epilogue combat=N/A is that violation. The recommendation neutralizes HL_0051 with an invented qualifier ("hard lines are only about chapters with a playable surface") not in the text.

3) RULING/GATE REWRITE PRESENTED AS DEFAULT. [SRC: CANON_AUDIT L63 Ruling 1] "L1 ... playable in EVERY chapter"; the harness R6 gate is "combat=Y in EVERY chapter, NEVER waivable, same class as unpopulated" [SRC: check_chapter_topography.py L35,124-135]. The Epilogue is a node in the sequence (chapter_number 78.0). Exempting it carves a fresh exception into a just-made ruling and a never-waivable hard-line gate — legitimate to propose, but framed as the canon-faithful default.

4) NOT A THROWAWAY. The Epilogue is a fully-authored spine node: own build slot ("Ch 77 → Epilogue → Final Range Review" [SRC: CASCADE_RUNBOOK L19]), entry file with three ending sections [SRC: L111], Range-Review coverage ("RR-FINAL audits ... plus CH_EPILOGUE" [SRC: L186]). This undercuts the "non-playable bookend, drop out of windows" framing.

NET: R4-binds + Prologue-IN survive and are worth ruling as recommended. The Epilogue combat=N/A / R6-exempt / OUT package does not survive as written — it over-reads path-invariance as non-playability, contradicts the Epilogue's affirmed hard-line status under HL_0051, and rewrites Ruling 1 / the never-waivable R6 while calling it the safe default. Material flaw: the Epilogue portion must be reframed from grounded-default to an EXPLICIT exception request (amending HL_0051 in the same breath), not applied. Disposition stays await_josh_confirm.

Canon-faithfulness flags raised:

• OVER-READ AS CANON: 'canon makes the Epilogue non-playable / combat not authored' is not what canon says — [SRC: HL_0045] and [SRC: RCM Epilogue block] state only that PATH-VARIANT epilogue content is not authored (path-invariance), never non-playability or combat-absence. Presented with [SRC] anchors as if grounded.
• HARD-LINE VIOLATION SMOOTHED OVER: the Epilogue is one of the 18 AFFIRMED hard-line nodes [SRC: CASCADE_RUNBOOK L18]; [SRC: HL_0051] says a hard-line chapter with no combat layer at all is a pillar violation ('combat absence is creatively fatal'). Setting Epilogue combat=N/A is that violation, neutralized only by an invented qualifier ('hard lines only apply to chapters with a playable surface') absent from HL_0051.
• RULING/GATE REWRITE AS DEFAULT: [SRC: CANON_AUDIT L63 Ruling 1] 'L1 ... playable in EVERY chapter' and R6 'combat=Y in EVERY chapter, NEVER waivable' [SRC: check_chapter_topography.py L35,124-135]. Exempting the Epilogue carves a new exception into a just-made ruling and a never-waivable hard-line gate, framed as the canon-faithful option rather than the exception-request it is.

Stronger option the adversary surfaced: A split disposition the recommendation collapsed into one all-or-nothing Option A. Stronger: (1) apply R4-all-five + Prologue-IN (survives cleanly); (2) for the Epilogue, DEFAULT to holding it inside the never-waivable R6 set (keep combat present, no N/A) UNLESS Josh affirmatively amends HL_0051 — rather than defaulting to N/A. This keeps the registry consistent with the Epilogue's affirmed hard-line status by default and forces the HL_0051 amendment to be explicit, mirroring how the audit treats the structurally-identical Ch 69 combat-layer question as [PROPOSAL - JOSH: write the exception into HL_0051] (CANON_AUDIT L77-78). Option A instead changes the data to combat=N/A first and flags the HL_0051 tension only as a minor sub-choice — risking a green gate sitting on an unamended hard-line contradiction.

Corrections to fold in:

• Split the package: RULE R4-all-five and Prologue-IN as recommended (they survive); do NOT bundle the Epilogue exemption as if it shared their canon footing.
• Re-pose the Epilogue as an EXPLICIT exception request, not a default: 'Epilogue is an affirmed hard-line node (CASCADE_RUNBOOK L18) rendering only the path-invariant unseen-reader frame (HL_0045, RCM Epilogue block); do you want to (a) amend HL_0051 to add an Epilogue-bookend exception and set combat=N/A, or (b) hold it to R6 like every other hard-line node?' Applying combat=N/A without the amendment leaves a live HL_0051 contradiction.
• Fix citations: replace 'canon makes the Epilogue non-playable' with 'canon makes the Epilogue path-invariant (HL_0045); playability/combat-presence is UNRULED and must be decided, not assumed.'
• If Josh exempts the Epilogue, write the amendment INTO HL_0051 (as the audit already anticipates for the structurally-identical Ch 69 case, CANON_AUDIT L77-78) so the hard line and the registry cell never contradict.

Apply actions. On Josh's confirm of Option A: (1) HARNESS — in harness/check_chapter_topography.py add an EPILOGUE bookend-exempt set (CH_EPILOGUE) that is excluded from the R6 combat-present gate and from R1/R2/R3 adjacency, while the Prologue remains in all windows under the default; keep R4 waivable. Update the module docstring's ledger note (L19-32) from 'Josh rules whether Pro/Epi count' to the ruling: 'Prologue IN windows (playable tutorial, RCM L6937); Epilogue OUT + R6-exempt (non-playable bookend, RCM L6968/HL_0045); R4 all-five binds, waivable, spine burns down 65 gaps.' (2) REGISTRY — in registries/'T0_Chapter_Index [ACTIVE v1.2]'/'T0_Chapter_Index DRAFT v1.1.csv' CH_EPILOGUE row: change mode_variety_check from 'combat=Y;puzzle=N;learning=N;character=Y;transformation=Y' to a bookend marker (proposed: 'combat=N/A;puzzle=N/A;learning=N/A;character=Y;transformation=Y' with an accompanying in-cell note, OR carry combat=N with the bookend-exempt flag so R6 does not fire) and set dominant_combat_mode from 'environmental' to a non-playable value consistent with 'no corpus, narrative-frame'; leave CH_PROLOGUE unchanged (playable, stays in windows). (3) DOCS — update docs/archive/LEDGER_BATCH1.md items §3 (R4 all-five-beats: mark RULED 'spine authors the missing beats; 65 waivers are a burn-down ledger'), the Epilogue-combat line (mark RULED 'Epilogue Ruling-1-EXEMPT; non-playable bookend'), and the Pro/Epi-window line (mark RULED 'Prologue IN, Epilogue OUT'); mirror into docs/PHASE0_LESSONS.md R4 note and docs/CASCADE_RUNBOOK.md pacing-gate contract. (4) Re-run the tests in the TEST field; commit with the ledger reference. NOTE — the exact Epilogue mode_variety_check encoding ('N/A' tokens vs combat=N+exempt-flag) is a small data-model sub-choice flagged in open_subflags for Josh to pick, since parse_variety currently accepts only Y|N.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Epilogue mode_variety_check ENCODING (Josh data-model pick): parse_variety currently accepts only Y|N per key [SRC: check_chapter_topography.py L85]. To make the Epilogue honestly non-combat without a red R6, choose: (a) add an 'N/A'/bookend token the parser recognizes for exempt nodes, or (b) keep combat=N and gate R6 by a CH_EPILOGUE exempt-set membership instead of the variety value. Recommend (b) — smaller parser change, keeps Y|N invariant.
• [DECISION NEEDED - JOSH] R4 waiver-format canon: confirm the per-chapter waiver carries a reason string + target ('all-five is the canonical target; this beat is [POPULATE→spine]') so the 65-count burn-down is auditable rather than a silent pass. Recommend requiring reason+target on every R4 waiver.
• [DECISION NEEDED - JOSH] Whether the Prologue, as a full-vril-ceiling tutorial, is graded peak/valley for R3 purposes (currently 'vril_management'/'standard' in the table) — its intensity label affects the Prologue→Ch1→Ch2 opening-window R3 result once Ch1/Ch2 are populated. Flagging for Josh since Prologue intensity is a pacing-canon call, not a data fix.

---

8. Four localized topography-pass conflicts: Ch 40-41 R1, Ch 37 flanking, Ch 57 fauna, Ch 41 Vimana numbering

• Key: per_chapter_conflicts_40_41_37_57_vimana Type: creative Disposition: await_josh_confirm

Options

• A — Resolve all four on canon (confirm CH_40→stealth; Ch 37 = pacing-flank not encounter; strike COMBAT L371 fauna clause for Ch 57; canonicalize Vimana six-count = modern chapters only, WGM outlier corrected) — (1) Confirm the 0.E re-pacing CH_40 puzzle_routed→stealth, dissolving the R1 {40,41} pair; ratify NO mode-level R1 exception. (2) Rule Ch 37's flanking-intensification is a pacing role (quiet flank ramping the Ch 38 peak) plus a low-intensity combat layer, not a combat ENCOUNTER at 37 — RCM 'no encounter' stands, consistent with Pillar 3. (3) Strike the 'person-to-animal with documented dangerous fauna' clause from COMBAT L371 for Ch 57 (companion edit); HL_0008/RCM 'no fauna combat' governs. (4) Canonicalize the six Vimana coordinates as the six modern full-reads (Ch 41=1, 44=2, 45=3, 47=4, 49=5, 50=6) with Ch 10/13 as unnumbered partial-read precursors; correct WGM's outlier 'Set 2 of 6' labels to 'Coordinate 1 of 6'.
• Canon support: (1) [SRC: docs/proposals/TOPOGRAPHY_REPACING.md L55-81] full R1/R2 math + rejection of mode-exception; [SRC: T0_Hard_Lines HL_0049] R1 exceptionless; [SRC: CVD §12.2 L664]; [SRC: T4_HoV HL_0104] its exception is scoped to operational TYPES (Ch40 FINANCIAL vs Ch41 EXTRACTIVE — already different), not combat modes. (2) [SRC: CVD Pillar 3 L182] flanking pairs '(Ch 37/39, 56/58, 68/70)' — Ch 37 flanks the Ch 38 peak; [SRC: RCM L4190] Ch 37 'no encounter... structural absence'; [SRC: Chapter_Index CH_37] already carries a dominant_combat_mode (not combat-absent). (3) [SRC: T0_Hard_Lines HL_0008] Ch 57 routes env+astral+vril only; [SRC: RCM L5465] 'no fauna combat' (twice); [SRC: docs/proposals/WGM_RECONCILIATION.md L51] WGM already rewritten to env/astral/vril; [SRC: COMBAT L371] is the lone outlier; CVD §17.1 hard line wins [SRC: CLAUDE.md authority hierarchy]. (4) [SRC: RCM L4433/4679/4738/4802/4939/5001] internally give Ch 41=1…Ch 50=6; [SRC: IPW L196-197] Ch 49=5, Ch 50=6 (agrees); [SRC: WGM L706] 'Ch 50... Coordinate 6 of 6' (only consistent if Ch 41=1); [SRC: WGM L564/L240, Languages L68] 'Set 2 of 6' is the outlier and internally contradicts WGM L706.
• B — Resolve the three grounded items on canon but DEFER the Ch 57 fauna strike to Josh as a hard-line touch — Same as A for conflicts 1, 2, 4, but treat editing COMBAT L371 (a §17.1 hard-line-adjacent combat clause) as Josh-exclusive since it narrows what combat is permitted on a maximum-protection chapter — recommend the strike but await confirm before touching COMBAT.
• Canon support: [SRC: CLAUDE.md] '§17 hard line' decisions and hard-line-status are Josh-exclusive; the Ch 57 clause sits at the §17.1 boundary. Against: the edit REMOVES a clause that contradicts an existing hard line rather than authoring new hard-line canon — it is a grounded correction, the same class already applied ~65× this session [SRC: CLAUDE.md Phase 0 note].
• C — Ratify Ch 40-41 as a mode-level R1 exception (extend HL_0104 to combat modes) and keep both puzzle_routed — Read HL_0104's Ch 40-41 anti-staleness exception as covering the dominant combat mode too, ratifying the shared puzzle_routed pair without re-pacing CH_40.
• Canon support: [SRC: T4_HoV HL_0104] does name a Ch 40-41 exception. Against (decisive): HL_0104's exception text is 'no two consecutive modern world chapters use same operational TYPE, except Ch 40-41' — scoped to antagonist operational type, not combat mode; [SRC: CVD §12.2 L664 / HL_0049] R1 on dominant combat type is exceptionless and CVD wins [SRC: CLAUDE.md]. Writing a new mode-level exception is canon invention contradicting a CVD rule sentence.

Recommendation — A — Resolve all four on canon (confirm CH_40→stealth; Ch 37 = pacing-flank not encounter; strike COMBAT L371 fauna clause for Ch 57; canonicalize Vimana six-count = modern chapters only, WGM outlier corrected)

How it fits the CVD. All four resolutions serve the CVD rather than bending it. (1) CH_40→stealth restores CVD §12.2/§16.2 R1 (no two consecutive chapters share dominant combat type) and Pillar 11 [SRC: CVD L664/L879], and preserves HL_0104's genuine operational-type anti-staleness at its own level. (2) Reading Ch 37 as a pacing-flank honors CVD §4.4 mountain-range pacing and Pillar 3's 'flanking chapter intensification' [SRC: CVD L182] literally — the intensification is the quiet-before-the-peak that makes the Ch 38 Azores convergence land — while Pillar 11's 'no chapter coasts / combat absence is fatal' is satisfied by the low-intensity combat layer the Chapter_Index already carries. (3) Striking COMBAT L371's fauna clause enforces CVD §17.1 maximum-protection at Ch 57 exactly as HL_0008/HL_0009 and RCM require; person-to-animal fauna combat on the songline/Dreaming chapter is precisely the 'shamanic-as-mystical-other / culture-as-antagonist' register the hard line forbids [SRC: HL_0005]. (4) The six-coordinate Vimana sequence is a numerical-adjacent canon fact; canonicalizing it to the six modern full-reads makes every T1 doc agree with the Ch 55 activation payoff [SRC: CVD §17.3 HL_0019; Vril §6.1 L328 'six-coordinate decoding sequence completion'].

How it builds canon. Closes four drift vectors before the spine cascade consumes these chapters. (1) Removes the last live R1 waiver in the modern-arc opening and gives Ch 39-43 a kinesthetic rotation (direct→stealth→moral-peak→environmental→puzzle) — the investigation gameplay becomes playable texture, unblocking the Ch 39-50 range for cascade. (2) Gives the cascade a clean rule for every hard-line flank: flanking chapters express intensification through pacing + a low-intensity combat layer, antagonist ENCOUNTER may still be structurally absent (remote-monitoring). (3) Makes Ch 57's combat-routing single-sourced (env+astral+vril) across COMBAT, WGM, RCM, HL — no downstream pipeline can pull 'fauna combat' into a §17.1 chapter. (4) Locks the Vimana coordinate ledger 41→44→45→47→49→50 so the Ch 55 activation and the six per-chapter unlock beats are consistent — a reusable spine anchor.

Why this option. Option A beats C because HL_0104's exception is textually scoped to antagonist operational TYPE, not combat mode, and R1 on dominant combat type carries no exception in CVD §12.2 — ratifying a mode-level exception would invent canon that contradicts a CVD rule sentence, and CVD wins every conflict. A beats B because the Ch 57 COMBAT L371 edit REMOVES a clause that contradicts an existing §17.1 hard line (HL_0008 + RCM's explicit 'no fauna combat') rather than authoring new hard-line canon — that is a grounded error-correction of exactly the class already applied ~65× this Phase-0 session, not a Josh-exclusive hard-line invention. Deferring it (B) leaves a live T1-vs-hard-line contradiction sitting on the maximum-protection chapter, which is the higher risk.

Three-layer fit.

• Surface (L1): L1: CH_40→stealth turns a repeated puzzle-boss beat into a supply-chain infiltration, and the fauna-clause strike means Ch 57 never asks the player to fight a goanna or crocodile — combat at Ch 57 stays environmental/astral/substrate. Ch 37 remains a quiet open-ocean traversal chapter whose low-intensity layer ramps the Ch 38 boss. All read complete on the surface.
• Mid (L2): L2: The Vimana six-coordinate ledger (Ch 41→50) is the investigatable through-line — a mid-depth player who decodes the six rock-art/geoglyph sites sees a coherent 1-of-6…6-of-6 arc culminating at Ch 55, with Ch 10/13 as the partial-read seeds; a mislabeled 'Set 2 of 6' would break that thread in the Grimoire/journal.
• Deep (L3): L3: The Ch 57 fauna-strike protects the deepest layer — the Dreaming/songline chapter surfaces convergent-tradition Pattern through recognition, not through the protagonist mastering or fighting the living substrate; person-to-animal combat there would collapse the recognition-not-mastery register the thesis depends on [SRC: HL_0094 'recognition, not mastery acquisition'].

Test (gate). After applying the CH_40 dominant_combat_mode puzzle_routed→stealth cell and removing the four topography waivers per docs/proposals/TOPOGRAPHY_REPACING.md, run python harness/run_gates.py (refresh docs/fidelity_baseline.json and docs/topography_baseline.json via --emit-baseline in the same commit). For the three doc edits, verify consistency anchors: COMBAT L371 no longer contains 'person-to-animal with documented dangerous fauna'; WGM L564/L240 + Languages L68 read 'Coordinate 1 of 6' consistent with WGM L706 'Ch 50... 6 of 6'; RCM L4433 unchanged (already 'Coordinate 1 of 6'). _Expected:_ run_gates.py exit 0 (green): R1 {40,41} dissolved, no new R1/R2 violation introduced at 39-40 (direct/stealth) or 40-41 (stealth/puzzle_routed); fidelity/topography baselines refreshed so the intended-edit set matches. Grep confirms COMBAT Ch 57 line lists only environmental + astral/dream-world + training-martial (fauna clause gone) and every Vimana '1 of 6' label agrees.

Adversarial review — HOLDS. I independently re-read every cited anchor. The four resolutions are canon-faithful in DIRECTION and Option A is correctly preferred over B and C. But the CONFLICT 3 apply text is materially under-inclusive — it would leave a live §17.1 contradiction — so the recommendation survives with a required correction, not clean.

WHAT HELD:

• CONFLICT 1 (CH_40 puzzle_routed→stealth): grounded. Verified CH_37=direct, CH_40=puzzle_routed, CH_41=puzzle_routed(peak) in the live Chapter_Index — the R1 {40,41} violation is real (both puzzle_routed). CVD §12.2 L664 "Two consecutive chapters never share dominant combat type" and HL_0049 both carry NO exception (verified verbatim). HL_0104's exception text reads "no two consecutive modern world chapters use same operational TYPE, except Ch 40-41" — textually scoped to operational TYPE, not combat mode. So Option C (mode-level exception) is correctly rejected as canon-invention against a CVD rule sentence, and RCM L4375's "inscription-ability reading of corporate internal-communications... Disruption" maps to COMBAT L327 "Stealth encounters: infiltration and extraction." Sound.
• CONFLICT 2 (Ch 37 = pacing-flank, no antagonist encounter): grounded. RCM L4190 verbatim "Antagonist operational form at Ch 37 is structural absence — no antagonist encounter at this chapter." CVD Pillar 3 L182 names the flanking pair "(Ch 37/39...)". Consistent with Pillar 3 + HL_0051 (combat LAYER present, ENCOUNTER absent). The recommendation honestly surfaces the residual tension (RCM L4190 "Vril substrate at Ch 37 not primary anchor") rather than hiding it.
• CONFLICT 4 (Vimana Ch 41=Coordinate 1 of 6): grounded and correctly flagged Josh-exclusive. IPW L196-197 verbatim: Ch 49 "Coordinate 5 of 6", Ch 50 "Coordinate 6 of 6". WGM L706 verbatim: "Ch 41 modern-era full read, Ch 50 ... Coordinate 6 of 6" — internally requires Ch 41=1. WGM L564/L240 "Set 2 of 6" are the internal outliers. The FINAL label choice is correctly held as Josh's numerical-canon call (open_subflag + await_josh_confirm).
• L3 claim: HL_0094 verbatim confirms "Ch 57 scene is recognition, not mastery acquisition" — the three-layer deep claim holds.
• Disposition await_josh_confirm is correctly conservative: CONFLICT 4 is numerical canon (Josh-exclusive), and Josh already held the CH_40 waiver "FOR JOSH." Not too aggressive.

WHAT BROKE (material, correctable):

CONFLICT 3's proposed edit strikes ONLY "person-to-animal with documented dangerous fauna, " from COMBAT L371, leaving the line reading "...combat through environmental encounters, TRAINING COMBAT WITH DOCUMENTED MARTIAL TRADITIONS, and astral or dream-world combat...". But every canonical source for Ch 57 sanctions ONLY environmental + astral + vril-substrate-defense: CVD §17.1 L933 ("Combat arrives through environmental and astral mode"), HL_0008/HL_0051 ("environmental, astral, and vril-substrate-defense modes"), and RCM L5465 (grep-confirmed: NO occurrence of "training combat" or "martial tradition" anywhere in the Ch 57 section; sanctions "environmental hazards plus astral mode plus vril-substrate-defense" and lists BOTH "no fauna combat" AND "no community combat" as §17.1 absolutes). "Training combat with documented martial traditions" is person-to-person combat with the community's own martial practitioners — it conflicts with §17.1 L923 "Never combat against community representatives... indigenous shamans... are inviolate" and HL_0003. The recommendation's own stated success criterion ("Ch 57 combat-routing single-sourced env+astral+vril across COMBAT/WGM/RCM/HL; no downstream pipeline can pull non-canonical combat into a §17.1 chapter") is NOT met by its edit, because a second non-canonical mode survives. Notably the companion WGM_RECONCILIATION L39 replacement gets this RIGHT ("environmental, astral, and vril-substrate-defense modes" — no martial training), so COMBAT and WGM would DISAGREE after the proposed edit — the opposite of the intended single-sourcing.

MINOR (cosmetic, not canon errors):

• Internal inconsistency: canon_support(2) says CH_37 "already carries a dominant_combat_mode (not combat-absent)" implying no cell edit, but apply_actions step 2 both labels itself "no file edit required" AND states "CH_37 dominant_combat_mode direct→vril_management." The live CSV shows CH_37 is still direct, so that edit is NOT yet applied — it is the separately-proposed TOPOGRAPHY_REPACING Hotspot-1 change, correctly bundled under the same Josh confirm. Substance is fine; the "no file edit required" label is wrong for step 2.

Canon-faithfulness flags raised:

• CONFLICT 3 apply text under-strikes: removing only the fauna clause from COMBAT L371 leaves 'training combat with documented martial traditions' as a Ch 57 combat mode, which is person-to-person combat against the community's own practitioners and conflicts with CVD §17.1 L923/L933 ('Never combat against community representatives'; 'Combat arrives through environmental and astral mode') and RCM L5465 'no community combat' absolute. The correct edit strikes/rewrites the WHOLE person-modes clause so the line reads only environmental + astral/dream-world (matching WGM_RECONCILIATION L39 and HL_0008/HL_0051). Leaving it creates a NEW COMBAT-vs-WGM disagreement, defeating the recommendation's own single-sourcing goal.

Stronger option the adversary surfaced: For CONFLICT 3 specifically, a stronger sub-option the recommendation missed: rewrite COMBAT L371's Ch 57 clause to the exact three-mode phrasing already drafted in WGM_RECONCILIATION L39 ('environmental, astral, and vril-substrate-defense modes'), rather than a minimal single-clause strike. This achieves the true single-sourcing across COMBAT/WGM/RCM/HL that Option A claims but its narrow edit does not deliver. Otherwise, no stronger top-level option exists — A correctly beats B (deferring the Ch 57 edit leaves a live T1-vs-hard-line contradiction on the maximum-protection chapter) and C (mode-level exception = canon invention against CVD §12.2/HL_0049).

Corrections to fold in:

• CONFLICT 3: broaden the COMBAT L371 edit. Do not strike only 'person-to-animal with documented dangerous fauna, '. Rewrite the Ch 57 combat-modes clause to list ONLY the three canonical modes — environmental encounters, astral/dream-world combat, and vril-substrate-defense — to match CVD §17.1 L933, HL_0008/HL_0051, RCM L5465, and the already-staged WGM_RECONCILIATION L39 replacement. Removing fauna but keeping 'training combat with documented martial traditions' leaves a second non-canonical person-combat mode on the maximum-protection chapter.
• apply_actions step 2: correct the 'no file edit required' label. The CH_37 direct→vril_management cell is NOT yet in the live Chapter_Index (verified: CH_37 currently = direct), so CONFLICT 2 does entail the TOPOGRAPHY_REPACING Hotspot-1 cell edit, not a pure rule ruling. Keep it bundled under the same Josh confirm but stop describing it as edit-free.
• CONFLICT 4 label wording: WGM L564/L240 use 'Set 2 of 6' as a cumulative-sequence-phase label, not strictly a per-chapter coordinate number; the recommendation's 'Coordinate 1 of 6' rewrite is defensible but the exact scheme (whether Ch 10/13 are counted) is correctly Josh's numerical-canon call — keep it in await_josh_confirm and do not auto-apply the label.

Apply actions. On Josh's confirm, apply as one gated commit (all edits exact-match, preserving trailing-whitespace bytes noted in WGM_RECONCILIATION L87):

1. CONFLICT 1 (grounded, Ruling-3/topography-delegated but bundle for sign-off): registries/T0_Chapter_Index [ACTIVE v1.2]/T0_Chapter_Index DRAFT v1.1.csv — CH_40 dominant_combat_mode puzzle_routed→stealth. Then docs/topography_baseline.json — remove R1 {CH_40,CH_41} waiver. Delete the 'R1 WAIVER' flag text from CH_40/CH_41 rows; KEEP the CH_41 Vimana [DECISION NEEDED] until step 4. (Josh already held this waiver 'FOR JOSH' — confirm.)

2. CONFLICT 2 (rule ruling, no file edit required): record Josh's ruling that Ch 37 flanking-intensification = pacing role + low-intensity combat layer, antagonist encounter structurally absent per RCM L4190; rewrite the CH_37 'intensification expression' [DECISION NEEDED] flag as resolved. CH_37 dominant_combat_mode direct→vril_management (already in the 0.E re-pacing, Hotspot 1).

3. CONFLICT 3 (grounded correction of a §17.1 contradiction): _source/01_Tier_1_Foundation/T1_Combat_System_Spec [ACTIVE v1.0].md L371 — strike 'person-to-animal with documented dangerous fauna, ' so the Ch 57 clause reads 'combat through environmental encounters, training combat with documented martial traditions, and astral or dream-world combat…'. (WGM Ch 57 rewrite already staged in WGM_RECONCILIATION L39/L51.)

4. CONFLICT 4 (numerical-canon — Josh-exclusive confirm): canonicalize six-coordinate scheme = modern full-reads Ch 41=1,44=2,45=3,47=4,49=5,50=6; Ch 10/13 = unnumbered partial-read precursors. Edit the outliers to 'Vimana Coordinate 1 of 6': WGM L564 and L240 ('Set 2 of 6'→'Coordinate 1 of 6', and fix the L564 parenthetical Ch 10/13 precursor labels); Languages L68 ('Set 2 of 6'→'Coordinate 1 of 6'); Threads L156 (normalize 'Ch 10 Set 1 partial + Ch 13 Set 2' precursor labels to 'Ch 10 + Ch 13 partial-read precursors'). RCM L4433 and IPW L196-197 already correct. Clear the CH_41 Vimana [DECISION NEEDED] flag.

Commit message carries the [SRC:] evidence per verification discipline; refresh both baselines with --emit-baseline in the same commit.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] CONFLICT 4 numbering scheme (Ch 41=Coordinate 1 of 6 with Ch 10/13 as unnumbered precursors) is numerical-adjacent canon — recommended on strong internal evidence (RCM+IPW+WGM-L706 all imply it) but the FINAL label choice ('Coordinate 1 of 6' vs a re-cut 'Set N of 6' scheme that counts Ch 10/13) is Josh's numerical-canon call.
• [DECISION NEEDED - JOSH] CONFLICT 2: whether Ch 37 must carry ANY combat layer at all, or may be the arc's one genuinely combat-absent flank, touches Pillar 11 'combat absence is creatively fatal' vs RCM 'structural absence' — recommended reading is low-intensity layer present (matches applied Chapter_Index), but the hard-line spirit call is Josh's.

---

9. Cross_Cultural_Link_Graph reward taxonomy + per-row assignment rule

• Key: CCLG_REWARD_TAXONOMY Type: creative Disposition: ratified_applied_2026-07-28 (Josh, sixteenth sitting — verbatim + consequences in docs/spine/DECISIONS_PENDING_JOSH.md; the ratified table is docs/proposals/REWARD_TAXONOMY_TABLE_2026-07-28.md; the 122-row apply landed 28o)

Options

• Option A — Instantiate CVD Pillar 5's own reward taxonomy as a closed enum, assign by a link_type→class rule — Add a reward_class column with the four canonical Pillar 5 payoff kinds verbatim: LORE (Grimoire/Thread-Map node + journal lore, granted on every link), WEAPON_ABILITY (cross-cultural weapon ability), SCHOOL_TECHNIQUE (school technique hybrid), INTUITION_TIER (intuition tier gain). Every link earns LORE; a deterministic rule maps the link_type/thread domain to the mechanical class (e.g. serpent/weapon-motif threads→WEAPON_ABILITY, school-anchored culture pairs→SCHOOL_TECHNIQUE, convergence/perception threads→INTUITION_TIER). Fill integrity_visibility with the same rule keyed to §9.2 bands; set component_anchor to CVD §3 Component 3 (already the proposed value). Flag only the genuinely ambiguous mechanical picks for Josh.
• Canon support: [SRC: CVD [ACTIVE v1.4] L196 Pillar 5 - "Linking earns lore reward + mechanical reward (cross-cultural weapon abilities, school technique hybrids, intuition tier gains)"] [SRC: CVD L123 - "Cross-Cultural Linking unlocks" as Layer-3 mechanical routing] [SRC: T1_Vril [ACTIVE v2.2] L170 - "Cross-Cultural Linking unlocks per CVD Pillar 5"] [SRC: T0_Schema_Dictionary [ACTIVE v1.3] §4.3 - authority owner is CVD §3 Component 3 + CVD Pillar 5; component_anchor proposed = CVD §3 Component 3]
• Option B — LORE-only default; leave all mechanical rewards [DECISION NEEDED - JOSH] — Set every row's reward to the LORE class (Grimoire node + journal lore) since that is the one payoff CVD grants unconditionally on every link, and leave WEAPON_ABILITY / SCHOOL_TECHNIQUE / INTUITION_TIER unassigned for Josh to author per link. Closes 145 rows at the safe floor; defers all mechanical-reward specifics.
• Canon support: [SRC: CVD L196 - lore reward is granted on every link unconditionally; mechanical reward is the variable part] [SRC: CVD L1042/L1164 - per-chapter Layer 2 Grimoire-link content is REQUIRED, so a lore node per link is always canon-safe] [SRC: CVD L197 executable test - reward without primary-source documentation = pillar violation, so a conservative default avoids over-claiming]
• Option C — Per-link bespoke rewards authored now — Author a specific mechanical reward for each of the 145 links individually (this exact weapon ability from this exact culture pair, etc.). Maximum specificity, no rule abstraction.
• Canon support: [SRC: CVD L196 names the reward kinds but not per-link specifics] [SRC: CLAUDE.md cardinal rule - Claude does not decide weapon/ability/school assignments; those are numerical/naming canon Josh authors] — canon does NOT support Claude authoring 145 bespoke mechanical payoffs; this would be invention at scale, the exact failure mode Known-issue drift warns against.

Recommendation — Option A — Instantiate CVD Pillar 5's own reward taxonomy as a closed enum, assign by a link_type→class rule

How it fits the CVD. The reward taxonomy is not invented — it is lifted verbatim from CVD Pillar 5 [SRC: CVD L196]: \"lore reward + mechanical reward (cross-cultural weapon abilities, school technique hybrids, intuition tier gains).\" That single sentence is the complete canonical payoff set; the four-value enum (LORE, WEAPON_ABILITY, SCHOOL_TECHNIQUE, INTUITION_TIER) is a one-to-one transcription. CVD L123 and T1_Vril L170 independently confirm \"Cross-Cultural Linking unlocks\" as a Layer-3 mechanical routing category, so the reward is real Layer-1 mechanics, never optional flavor. Pillar 5's executable test [SRC: CVD L197] — \"reward without primary-source documentation = pillar violation\" — is already satisfied structurally: every one of the 145 rows carries a populated evidence_basis SRC anchor and a scholarly_basis_classification enum [SRC: T0_Schema_Dictionary §4.3], so no row grants a reward without a scholarly basis. component_anchor resolves to CVD §3 Component 3 (cultural convergence), which the registry already proposes and the Schema Dictionary names as the authority owner [SRC: T0_Schema_Dictionary §4.3].

How it builds canon. Closes both open reward-relevant columns (component_anchor = CVD §3 Component 3; integrity_visibility keyed to the §9.2 five-band framework) and adds one typed reward_class column carrying the canonical enum, moving the flagship Pillar-5 registry from 0 usable reward cells to 145 assigned rows. It stands up the Grimoire payoff layer the cascade needs: the /spine-chapter cascade and Region Page Section 15 (Cross-Regional Connections) can now read a reward_class per link instead of a [DECISION NEEDED] stub [SRC: T0_Schema_Dictionary §4.3 - populate at Phase 6 region page build feedback through Section 15]. The link_type→reward_class rule is documented and auditable, so downstream cascade authoring inherits a defensible default rather than re-litigating 145 rewards. It unblocks Known-issue #4's \"empty flagship registry (Cross_Cultural_Link_Graph P5 = 0 rows)\" foundation-readiness gap.

Why this option. Option A beats B because B strands the mechanical reward — the \"highest-leverage innovation in the game\" per CVD L195 — as a permanent [DECISION NEEDED], leaving the cascade with nothing to route on; A gives a canon-grounded default that Josh can override per link rather than author from scratch. A beats C decisively on the cardinal rule: C would have Claude author 145 bespoke weapon/school/ability assignments, which are exactly the numerical/naming canon CLAUDE.md forbids Claude from deciding — that is invention at scale, the drift failure mode the repo exists to prevent. A threads the needle: the taxonomy is transcribed (not invented), the per-row class is assigned by a transparent rule (not a creative guess), and any row where the mechanical class is genuinely ambiguous is flagged for Josh rather than silently filled. The LORE class is 100% canon-safe on every row (per-chapter Grimoire link is CVD-required, L1164), so the floor is solid even before Josh rules on the mechanical picks.

Three-layer fit.

• Surface (L1): L1 SURFACE: each link fires a concrete Layer-1 mechanical payoff — a cross-cultural weapon ability charge, a school technique hybrid, or an intuition tier gain [SRC: CVD L196]. A surface-track player linking two serpent traditions gets a usable weapon ability and never needs to know the thesis; the reward is operational, not optional [SRC: CVD L123 Layer-3 mechanical routing per site].
• Mid (L2): L2 MID/DEPTH: every link also mints a Grimoire / Thread-Map lore node with primary-source attribution (the LORE class, granted on all 145 rows) [SRC: CVD L1140 - Grimoire links connect cross-regional patterns the player has surfaced; L50 - follow Grimoire links across cultures]. The depth player reads the scholarly basis (Frazer/Eliade/Campbell per row) behind why the two cultures link [SRC: CVD L196 scholarly basis required].
• Deep (L3): L3 DEEP: the reward taxonomy IS the thesis-as-gameplay-loop. Linking earns a reward precisely because 'we are all connected' — convergent traditions across disconnected cultures are the Pattern, surfaced by the player, never spoken [SRC: CVD L195 - the thesis becomes gameplay loop; L11/L1130 - surfaces through pattern, never named]. The integrity_visibility field ties the payoff's visibility to the player's integrity band, so the deep player's worldstate colors what the substrate reveals [SRC: CVD L965 five integrity bands; Schema Dictionary §4.4 Pillar-12 Layer-3 integrity-visible shift].

Test (consistency). Verify four anchors before applying: (1) the reward_class enum values are a verbatim subset of CVD L196's named payoffs (lore + weapon ability + school technique hybrid + intuition tier); (2) every one of the 145 rows already carries a non-empty evidence_basis SRC anchor and scholarly_basis_classification so no reward violates Pillar 5's executable test [SRC: CVD L197]; (3) component_anchor value CVD §3 Component 3 matches the Schema Dictionary §4.3 authority owner; (4) run harness/run_gates.py after the edit and refresh docs/fidelity_baseline.json via --emit-baseline in the same commit (the registry is a declared live-vs-frozen divergence per CLAUDE.md Phase-0 note). _Expected:_ reward_class enum ⊆ CVD L196 payoff names; 145/145 rows carry evidence_basis + scholarly_basis; component_anchor = CVD §3 Component 3; gates green with refreshed baseline.

Adversarial review — HOLDS. Refutation attempt: independently read every cited anchor. The core taxonomy grounding HELD — CVD L196 (Pillar 5) reads verbatim as cited ("lore reward + mechanical reward (cross-cultural weapon abilities, school technique hybrids, intuition tier gains)"), so the four-value enum {LORE, WEAPON_ABILITY, SCHOOL_TECHNIQUE, INTUITION_TIER} is a faithful transcription, not invention. L123 ("Cross-Cultural Linking unlocks") and T1_Vril L170 ("Cross-Cultural Linking unlocks per CVD Pillar 5") both confirm the reward is real Layer-1/3 mechanics, exactly as cited. Schema Dictionary §4.3 is the correct CCLG section; authority owner = CVD §3 Component 3 + Pillar 5; registry is scope-locked v0.1 placeholder. The live DRAFT CSV confirms 145 rows, and component_anchor + integrity_visibility are genuinely [DECISION NEEDED - JOSH] on all 145 rows. CVD §17.9 L1042 confirms a per-chapter Grimoire link is REQUIRED, so the LORE floor is canon-safe and canon-required — Option A's floor and Option B both rest on solid ground. Cardinal-rule discipline HELD: Option A explicitly refuses to author reward NAMES, flags the mechanical-reward RULE, the names, and the band cutoffs as Josh-exclusive (open_subflags), and sets disposition=await_josh_confirm. Option C is correctly rejected as invention-at-scale.

What BROKE — three flaws, one material: (1) MATERIAL FACTUAL ERROR. The cvd_fit and test both claim every row carries "a populated evidence_basis SRC anchor AND a scholarly_basis_classification enum [SRC: Schema Dictionary §4.3]," asserting Pillar 5's executable test (L197: reward without primary-source documentation = pillar violation) is "already satisfied structurally." I verified the live CSV header: there is NO scholarly_basis_classification column in the DRAFT CSV — that column exists only in the §4.3 schema SPEC, which the CSV does not implement (the CSV uses evidence_basis / link_type / link_strength, a different column set entirely). Worse, the evidence_basis SRC anchors on the rows I sampled point overwhelmingly to T1_Threads_22_Master_References (a T1 SECONDARY doc), not to primary sources (Frazer/Eliade/Campbell/primary field). So L197 is NOT structurally satisfied; the registry may in fact be non-compliant with L197 independent of reward_class. This is a false grounding claim. (2) SCHEMA MISMATCH unsurfaced: the DRAFT CSV and §4.3 spec have divergent column rosters; adding reward_class + bumping §4.3 without first reconciling the two is premature. (3) three_layer.deep cites "Schema Dictionary §4.4 Pillar-12 Layer-3 integrity-visible shift" as grounding for band-routed reward visibility — but §4.4 is T0_Creature_Roster and the integrity-visible shift in canon (Integrity_Paths §8.9.2) is scoped to FAUNA behavior, not link rewards. Extending it to reward visibility is analogical, not grounded; the claim overstates its footing (though the band cutoffs are correctly flagged Josh-exclusive).

Verdict: Option A beats B and C directionally and the disposition (await_josh_confirm, names/rule/cutoffs reserved for Josh) is appropriately conservative and canon-faithful. The recommendation survives as a directional call. But flaw #1 is a real defect that must be corrected before Josh rules: the claim that Pillar 5's L197 primary-source test is already satisfied is false, and the CCLG's evidence_basis currently cites a T1 secondary doc, not primary sources — that is arguably a pre-existing L197 exposure the dossier should surface, not smooth over.

Canon-faithfulness flags raised:

• cvd_fit/test claim 'scholarly_basis_classification enum' is populated on all 145 rows — FALSE: that column does not exist in the live DRAFT CSV (only in the §4.3 schema spec, which the CSV does not implement). Pillar 5 L197 executable test is therefore NOT structurally satisfied as claimed.
• evidence_basis SRC anchors on the sampled rows cite T1_Threads_22_Master_References (a T1 SECONDARY doc), not primary sources (Frazer/Eliade/Campbell/primary field) — CVD L197 requires primary-source documentation per link; the registry may be in pre-existing violation of L197 regardless of reward_class, which the dossier does not flag.
• three_layer.deep cites 'Schema Dictionary §4.4 Pillar-12 Layer-3 integrity-visible shift' for band-routed reward visibility, but §4.4 is T0_Creature_Roster and the canon integrity-visible shift (Integrity_Paths §8.9.2) is fauna-scoped, not reward-scoped — overstated grounding (band cutoffs are, however, correctly reserved as Josh-exclusive).

Stronger option the adversary surfaced: A stronger framing the dossier missed: split the decision into (i) the two genuinely-open, canon-safe column fills that require no mechanical-reward judgment — component_anchor = CVD §3 Component 3 (directly authority-owner-confirmed at Schema Dictionary §4.3) and reward_class=LORE floor (L1042-required) — apply-ready on Josh confirm; and (ii) the mechanical reward_class rule + the L197 primary-source-attestation gap, which should be a SEPARATE Josh decision after schema reconciliation. Option A bundles a clean, low-risk fill with a contestable game-design rule AND an unacknowledged L197 compliance problem, raising the odds Josh bounces the whole thing. Decoupling lets the safe 80% land now and isolates the real open question.

Corrections to fold in:

• Delete the claim that Pillar 5's L197 executable test is 'already satisfied structurally' by a scholarly_basis_classification enum. That column is not in the live CSV. Rewrite the test to (a) reconcile the DRAFT CSV column roster with §4.3 spec first, and (b) verify evidence_basis actually cites PRIMARY sources per L197 — currently it cites T1_Threads_22, a secondary doc.
• Surface, do not smooth over, the pre-existing L197 exposure: the CCLG's 145 rows currently ground links on a T1 secondary reference, not primary-source documentation. Flag this to Josh as a compliance gap that the reward taxonomy sits on top of but does not fix.
• Reconcile the DRAFT CSV schema with Schema Dictionary §4.3 (they have entirely different column names) BEFORE adding reward_class and bumping §4.3; the apply_actions currently assumes the two already match.
• Correct three_layer.deep: cite the fauna-scoped precedent (Integrity_Paths §8.9.2 / §4.4 Creature_Roster) as an ANALOGY for band-routed reward visibility, not as direct grounding; keep the band cutoffs flagged Josh-exclusive.

Apply actions. On Josh's confirm: (1) Add a typed reward_class column to registries/\"T0_Cross_Cultural_Link_Graph [DRAFT v0.1]\"/Sheet1.csv with the closed enum {LORE, WEAPON_ABILITY, SCHOOL_TECHNIQUE, INTUITION_TIER} (MINOR schema bump per Schema Dictionary §2.3; also add the enum to T0_Schema_Dictionary §4.3 column roster). (2) Set reward_class=LORE on all 145 rows (canon-safe floor) and additionally assign the mechanical class by rule: weapon/serpent/bull-sacrifice/guardian motif threads (Thread 2, weaving 13/15) → WEAPON_ABILITY; school-anchored culture pairs whose chapters are Pillar-4 school anchors (Ch 15/17/18/21/23/24/27/29/47/66-69 per CVD L189) → SCHOOL_TECHNIQUE; vril-parallel/convergence/perception threads (Thread 1, catalog-convergence, catalog-membership, parallel-independent) → INTUITION_TIER. (3) Replace component_anchor [DECISION NEEDED] with 'CVD §3 Component 3' on all 145 rows. (4) Fill integrity_visibility with a §9.2-band value using the same band framework (default: visible from L1 Truly Good through L3 Drifting; Evil/Pure-Evil bands see the corrupted-substrate variant per Pillar-12 Layer-3 integrity-visible shift). (5) Leave [DECISION NEEDED - JOSH] on any row where the mechanical class is genuinely ambiguous (list them for Josh). (6) Run harness/run_gates.py and refresh docs/fidelity_baseline.json with --emit-baseline in the same commit. Do NOT author bespoke per-link weapon/school/ability NAMES — those remain Josh-exclusive.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Josh-exclusive: the link_type→mechanical-reward-class assignment RULE (which thread domains map to WEAPON_ABILITY vs SCHOOL_TECHNIQUE vs INTUITION_TIER) is a game-design call; Claude's proposed rule is grounded in CVD Pillar-4 school-anchor chapters and thread-motif domains but the final mapping is Josh's to confirm.
• [DECISION NEEDED - JOSH] Josh-exclusive: any per-link bespoke reward NAME (specific weapon ability, specific school technique hybrid, specific intuition-tier magnitude) is numerical/naming canon and must not be Claude-authored; only the reward_CLASS is proposed here.
• [DECISION NEEDED - JOSH] integrity_visibility band-routing default (L1-L3 see canonical form; L4-L5 see corrupted-substrate variant) is proposed from Pillar-12 Layer-3 precedent but the exact band cutoffs per reward are Josh's to rule.

---

10. Thread_Grid: 35 anchor-gap cells + L145 Thread-2 contradiction + Thread 4/6 "every chapter" invariant

• Key: threadgrid_35_gaps_L145_invariant Type: creative Disposition: await_josh_confirm

Options

• Option A — Author a fresh Josh anchor for every one of the 35 gap cells (fill the grid to a literal one-anchor-per-cell state) — Treat the Threads §Purpose statement 'Every chapter advances Threads 1-6' [SRC: THREADS L53, L57] as a literal per-cell mandate and have Josh author 35 new per-chapter anchors (Hebrew ruach for Ch 18, Tibetan rlung for Ch 20, Mongol hiimori for Ch 22, a fairy-realm serpent placement for each of Ch 70-75, a Thread 6 rising-phase onset for Ch 70-73, etc.) so no Major-Thread cell is ever empty.
• Canon support: FOR: Threads §Purpose 'Every chapter advances Threads 1-6' and §Typology 'Threads 1-6 ... universally tracked in every chapter' [SRC: THREADS L47, L53, L57]. AGAINST (decisive): the domain doc's OWN per-thread catalogs already name anchors for far fewer than 79 chapters — Thread 4 catalog skips Ch 3/7/9/16/20/22/25 and the whole 51-63 stretch, Thread 6 catalog skips Ch 2-3/5/8/10/12-13/16/18/20-22/25/33-41/43-46/48-51/53-63/65-69 [SRC: THREADS L267-302 Thread 4 catalog; L349-372 Thread 6 catalog; grid THREAD_04 has cells for 38 chapters, THREAD_06 for 30 of 79]. §8 explicitly authors negative anchors: 'Ch 33 ... no Thread 2 anchor canonical', 'Ch 34-37 ... no anchor', Ch 57 Thread 3 'ABSENT-by-design', Ch 74/75 'No ... Thread direct anchors' [SRC: THREADS L144-145, L167, and grid rows 137-142 quoting §8]. CVD §18.3 forbids fabricated substrate; inventing 35 anchors to satisfy a framework phrase is exactly the invented-canon-at-scale failure CLAUDE.md's cardinal rule guards against.
• Option B — Triage by family: NOT_PRESENT the genuine dark cells, fix Chapter_Index for L145, author only the 5 canon-supported derivations, reframe the invariant as framework-continuity (RECOMMENDED) — Route each of the 35 cells to the resolution its own canon supports. ~30 cells where canon names no anchor (and §8 often affirmatively excludes one) become progression_stage=NOT_PRESENT with the negation text preserved — using the convention that ALREADY EXISTS in the grid. 5 cells carry a canon-grounded derivation Josh can confirm (Thread 5 Ch 12 iroko localization; Thread 5 Ch 70 RCM transit-register; Thread 1 Ch 14/32 vril-infrastructure derivations; Thread 1 Ch 59 Baka re-read). The L145 contradiction resolves by fixing the downstream registry (Chapter_Index T02:A for Ch 33-37 is wrong; the domain-owner doc wins). The 'every chapter' invariant is documented as framework-level thematic continuity, not a per-cell fill mandate.
• Canon support: FOR: NOT_PRESENT already exists as a canonical progression_stage (7 cells in-grid) [SRC: grid rows 201/213-218; distinct-value count 7×NOT_PRESENT]; Ch 57 Thread 3 'ABSENT-by-design' is the clean precedent [SRC: THREADS L167 §8; grid row 201]. §8 affirmatively excludes anchors at Ch 70/74/75 ('No ... Thread direct anchors') [SRC: THREADS §8, grid rows 137/141/142]. Threads doc is the domain owner and wins over the T0 registry per authority hierarchy [SRC: THREADS L17-19; CLAUDE.md authority hierarchy]. The 5 derivations each cite specific canon (iroko=Yoruba, Ch 12=Yoruba chapter [SRC: THREADS L74, L319, grid row 268]; RCM Ch 70 Thread-5 transit-register [SRC: grid row 294 quoting RCM Ch 70]). AGAINST: none material — this is the disposition each cell's own reasoning already reaches.
• Option C — Blanket rule: every gap cell is dark; drop/blank all 35 rows and take no derivations — Rule uniformly that a gap = absence: set all 35 to NOT_PRESENT (or drop the rows), author nothing, and fix Chapter_Index to match. Simplest and fully invention-free, but discards the handful of cells where canon DOES support a localizable anchor and would leave real thematic beats unrecorded.
• Canon support: FOR: maximal fidelity to do-not-invent; every cell defensible as 'canon names no discrete anchor here.' AGAINST: over-corrects — Thread 5 Ch 70 has an explicit RCM Thread-5 anchor ('World Tree root system dimensional transit from Maatherion Ch 69 to fairy realm Ch 70') [SRC: grid row 294 quoting RCM Ch 70]; Thread 5 Ch 12 iroko is catalogued in the very range entry 'Ch 11-13: Iroko + baobab' and iroko is Yoruba = the Ch 12 culture [SRC: THREADS L319; grid row 268]. Blanking these loses canon that exists, and leaves the Thread 5 Ch 70 vs §8 conflict unreconciled.

Recommendation — Option B — Triage by family: NOT_PRESENT the 30 genuine dark cells + reconcile L145 by fixing Chapter_Index (not the domain doc) + author only the 5 canon-supported derivations, and reframe the "every chapter" invariant as framework-continuity

How it fits the CVD. The threads exist to carry the cross-cultural thesis 'in the mechanical structure of game content rather than declared in dialogue' [SRC: THREADS L51-53] — the Layer-3 convergent-Pattern engine. CVD §18.3 forbids fabricated/asserted substrate; a thread must surface through real cultural material, never be manufactured to fill a slot [SRC: CVD §18.2-18.4, L1134-1144]. Option B honors this: it records a thread as live only where a culture's actual tradition anchors it, and marks honest absence where the arc's structure (open-ocean crossing Ch 34-37; finale chapters where §8 states 'No Thread direct anchors') gives the thread nothing to attach to. Fabricating a Hebrew/Tibetan/Mongol vital-force term (Option A) would be Layer-2 substrate the doc never documents — a §18.3 violation and a direct breach of CLAUDE.md's do-not-invent rule. Reconciling L145 in favor of the domain doc respects the CVD authority hierarchy (T1 domain doc > T0 registry).

How it builds canon. Stands up the Thread_Grid as a trustworthy Layer-3 retrieval layer for the Phase 6 spine cascade: every Major-Thread cell becomes machine-readable as either a grounded anchor, a confirmed derivation, or an explicit NOT_PRESENT — no silent gaps, no fabricated content. It formalizes NOT_PRESENT as the canonical dark-cell convention (already used 7×, now applied consistently), which the cascade's per-chapter thread-advance step can query directly. It repairs a real data-model bug (Chapter_Index T02:A for Ch 33-37 contradicting the domain owner), preventing that error from propagating into spine v2.4. And it converts a soft framework phrase ('every chapter advances Threads 1-6') into a precise, defensible statement (framework-continuity, not per-cell fill) that future audits won't re-flag.

Why this option. Option A is refuted by the domain doc's own catalogs: Thread 4 names anchors for ~38 chapters and Thread 6 for ~30 of 79, and §8 affirmatively AUTHORS absence ('no Thread 2 anchor canonical', 'ABSENT-by-design', 'No Thread direct anchors'). So the 'every chapter' phrase cannot mean literal per-cell fill — treating it as such forces 35 inventions and breaks the cardinal rule. Option C is invention-free but throws away the 5 cells where canon genuinely supports a localized anchor (notably Thread 5 Ch 70, which has an explicit RCM Thread-5 transit anchor, and Thread 5 Ch 12 iroko). Option B is the only route that matches each cell to the disposition its own grounded reasoning already reached: absence where canon is silent, a Josh-confirmable derivation where canon points at one, and a registry fix (not a canon edit) for the L145 contradiction. It is also the least-invention path that still preserves every real beat.

Three-layer fit.

• Surface (L1): L1 SURFACE: invisible. Thread anchors are authoring-substrate; a surface player never sees a grid cell. Whether Ch 20 carries a Tibetan vril-term or is marked NOT_PRESENT changes nothing about the playable Silk Road chapter.
• Mid (L2): L2 MID/DEPTH: this is where the decision bites. The depth player reads convergence through documented traditions (journal/Grimoire cross-cultural links). A FABRICATED anchor (Option A's invented ruach/rlung/hiimori) would be false substrate a depth player could catch as unsupported — corrosive to the exact trust Layer 2 depends on. Option B only surfaces anchors backed by real tradition, and its 5 derivations are all attested (iroko=Yoruba, RCM's own World-Tree transit text), so the depth layer stays honest.
• Deep (L3): L3 DEEP: the thesis is the convergence-Pattern, never spoken, surfaced only where cultures independently attest the same phenomenon [SRC: CVD §18.2 L1134]. A dark cell does NOT weaken the Pattern — the Pattern is cumulative across 77 chapters, not per-chapter [SRC: THREADS L53 'Cumulative thread progression across 77 chapters']. Recording honest absence keeps the convergence evidentiary rather than manufactured; a padded grid would dilute the very signal Layer 3 is built from.

Test (consistency). Verify five anchors before applying: (1) grid distinct progression_stage values already include NOT_PRESENT (7 cells) — confirms the convention is pre-existing, not invented [grep ',NOT_PRESENT,' in the grid CSV]. (2) Chapter_Index CH_33..CH_37 each carry T02:A while THREADS L144-145 states 'Ch 33 ... no Thread 2 anchor canonical' and 'Ch 34-37 ... no anchor' — confirm the registry is the wrong side. (3) THREADS Thread-4 catalog (L267-302) contains no Ch 24 entry (Ch 23 Deucalion → Ch 26 Slavic) and Bergelmir is homed at Ch 27 — confirms Thread 4/Ch 24 = NOT_PRESENT. (4) THREADS L319 range entry 'Ch 11-13: Iroko + baobab' + L74 'ase (Yoruba) Ch 12' — confirms iroko localizes to the Yoruba chapter (Ch 12). (5) After edits, re-run the Thread_Grid fidelity/anchor gate and confirm zero anchor_gap_josh flags remain and no cell has empty content_summary. _Expected:_ (1) 7 NOT_PRESENT cells present. (2) Chapter_Index is the erroneous side; domain doc wins. (3) no Ch 24 Thread-4 catalog entry. (4) iroko is Yoruba/Ch 12. (5) gate green, zero remaining anchor_gap_josh flags, every Major-Thread cell resolved to anchor|derivation|NOT_PRESENT.

Adversarial review — HOLDS. I attempted to refute Option B on all six attack surfaces and independently verified every load-bearing anchor against the ACTIVE canon. The recommendation survives; the refutation attempts strengthened rather than broke it.

WHAT I VERIFIED (anchors check out):

1. L145 contradiction is REAL and the fix direction is correct. Chapter_Index rows CH_33..CH_37 all carry T02:A (verified in the CSV thread_anchors_complete column), while THREADS L144 ('Ch 33 Atlantic Approach no Thread 2 anchor canonical') and L145 ('Ch 34-37 Atlantic-crossing Thread 2 no anchor per open-ocean chapter-type') author explicit absence. THREADS is the domain owner; per the CVD authority hierarchy (T1 domain > T0 registry) the registry is the erroneous side. Fix = registry, not the doc. Confirmed.

2. The grid's THREAD_02 CH_33-37 cells are mislabeled progression_stage=ANCHOR while their content_summary is the NEGATION text — a genuine data bug; normalizing to NOT_PRESENT is correct.

3. NOT_PRESENT is a pre-existing convention (verified exactly 7 cells: THREAD_03 CH_57 'ABSENT-by-design' and CH_69-74), so Option B invents no new convention.

4. Every Group-1 NOT_PRESENT cell is grounded in the domain catalog: Thread 1 §2 (L69-96) skips Ch 16/18/20/22/49 (no vital-force term); Thread 2 §2-3 'Continental revisits (Ch 59-65)' (L171-173) lists ONLY Ch 64, and Final Act (L175-180) anchors Ch 66/68/69 and skips Ch 67/70-73; Thread 5 (L310-341) skips Ch 71/73/74/75; Thread 6 (L349-372) skips Ch 70-73. All verified line-by-line.

5. The 5 derivation cells each carry a live [DECISION NEEDED - JOSH] flag whose own reasoning already reaches the recommended disposition; iroko is confirmed range-level ('Ch 11-13', L319) hence genuinely a Josh-confirm derivation, not a fact.

REFUTATION ATTEMPTS THAT FAILED:

• 'Option A is grounded by the every-chapter mandate' — FALSE. Refuted by the domain doc's own catalogs AND, more decisively, by CVD itself: §3 Component 4 (L83) writes 'every chapter (Velheim presence except Ch 57 Australia)' — the constitutional authority models its own 'every chapter' phrase as thematic-continuity admitting per-chapter exceptions. This is a HIGHER-authority anchor than the recommendation even cited, and it kills Option A's premise. The invented terms Option A would author (Hebrew ruach Ch 18, Tibetan rlung Ch 20, Mongol hiimori Ch 22) appear nowhere in the Thread 1 catalog — pure fabrication, a direct cardinal-rule (do-not-invent) and CVD §18.3 (no fabricated substrate) violation.
• 'CVD L95 every chapter carries Layer 3 mechanical routing makes a NOT_PRESENT Major-Thread cell a hard-line violation' — FALSE. L95 governs Layer-3 routing through sites/NPCs/phenomena, not a discrete-anchor-per-thread-per-chapter requirement; the thread-owning doc authors absence explicitly and CVD writes its own 'except Ch 57' exception. No §17 hard line is smoothed over.
• 'Disposition too aggressive' — FALSE. disposition=await_josh_confirm; all apply_actions gated 'On Josh's confirm'; the Thread 5 Ch 70 §8-vs-RCM conflict, the Thread 1 Ch 14/32/59 category stretch, and the iroko localization are all surfaced as open sub-flags rather than silently applied; and it does NOT auto-patch the ACTIVE _source THREADS §8 doc. Correct discipline throughout.

Three-layer claims hold: the decision is L1-invisible (authoring substrate), bites at L2 (a fabricated anchor would be catchable false substrate), and a dark cell does not weaken the cumulative L3 Pattern (THREADS L53 'Cumulative thread progression across 77 chapters'). All accurate.

No material flaw changes the verdict. Option B is the least-invention path that preserves every real beat, fixes a genuine registry bug on the correct side of the hierarchy, and defers every judgment call to Josh.

Stronger option the adversary surfaced: none — Option B already routes each cell to the disposition its own grounded reasoning reaches; A over-invents, C over-corrects. The only improvement is evidentiary (add the CVD §3 Component 4 'except Ch 57' anchor), not a different option.

Corrections to fold in:

• Strengthen the FOR case with a higher-authority anchor the recommendation missed: CVD §3 Component 4 (L83) itself writes 'every chapter (Velheim presence except Ch 57 Australia)' — the CVD, the constitutional authority, models an 'every chapter' phrase as thematic-continuity that admits per-chapter exceptions. This is stronger than the Threads-catalog argument alone and should anchor the invariant-reframe (framework-continuity, not per-cell fill).
• Minor precision note for apply time: the Group-1 list correctly EXCLUDES Thread 2 CH_64 (L173 gives Ch 64 a real submerged-civilization serpent anchor) and correctly EXCLUDES Thread 6 CH_74/75 (L371 gives them the protagonist's-own-arc dying-rising anchor). Preserve these exclusions verbatim — do not let a 'blanket the 60s/70s' mechanical pass sweep CH_64 (Thread 2) or CH_74/75 (Thread 6) into NOT_PRESENT.
• The THREAD_05 CH_70 reconciliation implies a downstream edit to the ACTIVE _source THREADS §8 index (which currently says 'No other Major or Minor Thread direct anchors'). Keep this strictly as a Josh-confirm sub-flag; do NOT edit _source. The grid cell can adopt the RCM transit text with extensions.josh_authored=true, but the §8-vs-RCM authority call is Josh's.

Apply actions. On Josh's confirm, apply in the DRAFT grid + Chapter_Index (all in-repo, no _source edits):

GROUP 1 — NOT_PRESENT (30 cells; grounded absence, apply mechanically): set progression_stage=NOT_PRESENT, keep the negation content_summary, drop the anchor_gap_josh flag. Cells: THREAD_01 CH_16, CH_18, CH_20, CH_22, CH_49 (canon names no vital-force term; §8 omits Thread 1); THREAD_02 CH_60, CH_61, CH_62, CH_63, CH_65, CH_67, CH_70, CH_71, CH_72, CH_73, CH_74, CH_75 (§2-3 range-level only + §8 explicit 'No Thread direct anchors' at 70/74/75); THREAD_03 CH_01 (catalog opens Ch 2); THREAD_04 CH_24 (catalog skips Ch 24; Bergelmir homed Ch 27); THREAD_05 CH_71, CH_73, CH_74, CH_75 (§8 excludes; range-level only); THREAD_06 CH_70, CH_71, CH_72, CH_73 (§8 anchors Thread 3 only; Weaving-9 framework continuation is not a per-chapter anchor). ALSO normalize the 5 existing 'precedent' cells THREAD_02 CH_33-37 from ANCHOR→NOT_PRESENT for convention consistency.

GROUP 2 — DERIVE, Josh-confirm text (5 cells; canon points at an anchor, mark extensions.josh_authored=true): THREAD_05 CH_70 → adopt the RCM Ch 70 Thread-5 transit-register text ('World Tree root system dimensional transit from Maatherion Ch 69 to fairy realm Ch 70') and note §8 Ch 70 needs reconciliation; THREAD_05 CH_12 → iroko localized to Ch 12 (Yoruba); THREAD_01 CH_14 → derive from the WGM §5.2 obelisk vril-infrastructure anchor; THREAD_01 CH_32 → derive from the RCM Ch 32 vril-perception flood-memory secondary beat; THREAD_01 CH_59 → derive from the Trade-4 Baka re-read of Thread 1's Ch 11 relational-field depth.

GROUP 3 — DROP/blank (2 cells, stale pre-v2.2 text): THREAD_11 CH_45, CH_46 → set NOT_PRESENT (Thread 11 chain is Ch 43→47; the 'Ch 43-46 Brer Rabbit' blurb exists nowhere in _source).

REMAINING GAP CELL THREAD_03 CH_10 → NOT_PRESENT (range-carrier Ch 9-13; no Ch 10 antagonist-family content) unless Josh authors a Ch 10 Thread-3 anchor.

L145 FIX: in Chapter_Index rows CH_33..CH_37, change the T02 code from T02:A to reflect no-anchor (either drop the T02 token or set the registry's no-anchor code), so the registry matches the domain owner THREADS L144-145. Record in the Living Revision Log.

INVARIANT DOC-NOTE: append a note in the Threads doc's downstream tracking / cascade runbook that 'Every chapter advances Threads 1-6' is framework-continuity (cumulative Pattern across 77 ch), NOT a per-cell fill mandate — Threads 4/6 canonically anchor ~30-38 discrete chapters; NOT_PRESENT cells are canon-compliant.

Then re-run the Thread_Grid gate; expect zero anchor_gap_josh flags.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Thread 5 Ch 12 iroko localization (Ch 11 vs Ch 12): canon has iroko in the 'Ch 11-13' range and iroko is Yoruba=Ch 12, but T1_Story_Spine gives Ch 11 the Ituri canopy-axis form — Josh confirms whether iroko anchors at Ch 12 as recommended [SRC: THREADS L319, L74; grid row 268].
• [DECISION NEEDED - JOSH] Thread 5 Ch 70 §8-vs-RCM reconciliation: RCM Ch 70 carries a Thread-5 transit anchor but THREADS §8 Ch 70 says 'No other Major or Minor Thread direct anchors' [SRC: grid row 294]. Recommend adopting RCM and patching §8; Josh confirms which source is authoritative for this cell.
• [DECISION NEEDED - JOSH] Thread 1 Ch 14/32/59 derivations reclassify vril-INFRASTRUCTURE / Trade-labeled beats as Thread-1 (Vril Parallels) anchors — a small category stretch (parallel-TERM vs vril-field). Josh confirms whether these count as Thread 1 anchors or should be NOT_PRESENT.
• [DECISION NEEDED - JOSH] Thread 3 Ch 10 and Ch 01: recommend NOT_PRESENT (catalog opens Ch 2; Ch 9-13 is a range-carrier with no Ch 10 content), but if Josh wants Thread 3 literally present every mortal-world chapter he would author Ch 10 (and possibly Ch 1) anchors instead.
• [DECISION NEEDED - JOSH] Whether to author fairy-realm serpent (Thread 2) and rising-phase (Thread 6) placements for Ch 70-75 at all: canon gives only range-level 'Ch 70-77 fairy realm's own serpent tradition' with §8 explicitly excluding per-chapter anchors — recommend NOT_PRESENT, but this is Josh-exclusive finale canon if he wants discrete placements.

---

11. L3 seeds Ch 34/35/36 + the per-chapter Layer-3 storage convention (HL_0047)

• Key: L3_SEEDS_CH34_36_AND_HL0047_STORAGE Type: creative Disposition: await_josh_confirm

Options

• A — Adopt proposal as drafted: spine v2.4 brief is the standing per-chapter L3 carrier; RCM appends + three §18.5 registries + Chapter_Index.time_fluctuation_range are the pre-cascade landing — Confirm the storage convention exactly as docs/proposals/L3_SEEDS.md drafts it. Per-chapter L3 anchor FIELD (orphaned when T2 retired) is re-homed to the spine v2.4 per-chapter brief; the RCM Verified-canon appends carry the prose seed now; the three CVD-named registries (Vril_Site, Cross_Cultural_Link_Graph, Inscription_Spine) carry the cross-regional pattern weight; Chapter_Index.time_fluctuation_range holds the time-fluctuation environmental-routing datum. Apply the three RCM appends + VS_CH34_001 + the three time_fluctuation_range cells now (gated by the critic), defer the Josh-exclusive row specifics.
• Canon support: [SRC: CVD "...[ACTIVE v1.4].md" §18.4 L1162 "Every chapter has at least one Layer 3 mechanical event... T2 Chapter Design briefs carry the Layer 3 anchor field" — the field's original home] [SRC: CVD §18.5 L1178 "The quest's Layer 3 anchor point in T0_Inscription_Spine, T0_Vril_Site_Registry, or T0_Cross_Cultural_Link_Graph carries the cross-regional pattern weight" — the three registry stores] [SRC: SSBL "...[ACTIVE v3.2].md" §9.7 Protocol 7 L668 "does this chapter's mechanical routing surface the metaphysical-as-real substrate per Component 7" — the spine consumes the per-chapter L3 check] [SRC: CLAUDE.md Known-issue #4 Ruling "spine v2.4 = ENHANCE the v1.0 baseline" — the brief is the standing cascade carrier]
• B — Registries-only carrier: drop Chapter_Index.time_fluctuation_range from the L3-carrier set; L3 anchor field lives ONLY in the spine v2.4 brief + the three §18.5 registries — Same as A but excludes time_fluctuation_range from the standing L3 convention. Rationale: the column has no Schema-Dictionary definition and its 40 existing populated values are era/place SETTING windows (CH_09 "Aksum; Lalibela", CH_12 "Akan territory"), not L3-mechanic descriptors. Populate the CH_34/35/36 cells as a data-completeness pass (they are legitimately empty), but do not canonize the column as the per-chapter L3 carrier. The L3 field is carried by the spine brief; the three CVD-named registries carry pattern weight.
• Canon support: [SRC: T0_Schema_Dictionary "...[ACTIVE v1.3].md" — grep for time_fluctuation_range returns NO definition; the column is unschematized] [SRC: registries/T0_Chapter_Index.../...DRAFT v1.1.csv col 8 — populated values are era/place windows e.g. CH_09 "Aksum; Lalibela; Ras Dashen", CH_12 "Akan territory (modern Ghana)", NOT L3-mechanic strings] [SRC: CVD §18.5 L1178 names exactly three registries; time_fluctuation_range is not among them] [SRC: T1_Vril §11.1 L319 "Time fluctuation is canonically a vril-instability phenomenon" — time fluctuation IS L3-adjacent, so the cell content is defensible even if the column is not THE L3 carrier]
• C — Defer everything to the cascade: author no pre-cascade seeds; let the spine v2.4 /spine-chapter cascade author Ch 34/35/36 L3 in-brief when it reaches those chapters — Do not land any RCM appends or registry rows now. Treat HL_0047 per-chapter compliance as a cascade deliverable authored inside each chapter's spine brief at build time. Rationale: the spine cascade is the canonical consumer of the L3 check (SSBL §9.7) and re-authors chapters anyway; pre-seeding risks drift the cascade would have to reconcile.
• Canon support: [SRC: SSBL §9.7 Protocol 7 L668 — the spine cascade is where the per-chapter three-layer verification canonically runs] [SRC: CLAUDE.md "Phase 6 (Story Spine cascade): NOT started" — the cascade is the main work ahead] [SRC: CVD §18.4 L1162 — the L3 anchor FIELD is a per-chapter-brief artifact, and briefs are a cascade output] [Against: CLAUDE.md Known-issue #4 "foundation-readiness gaps should be closed before/early in the cascade" — the audit explicitly wants these closed before the cascade, not deferred into it]

Recommendation — A — Adopt proposal as drafted, with one refinement absorbed from Option B

How it fits the CVD. The storage convention is the only reading that keeps HL_0047 satisfiable after T2 retirement without inventing canon. CVD §18.4 [L1162] created a per-chapter "Layer 3 anchor field" and parked it in T2 Chapter Design briefs; Phase 5.5 deleted that tier, orphaning the field. CVD §18.5 [L1178] independently names three registry stores (T0_Inscription_Spine, T0_Vril_Site_Registry, T0_Cross_Cultural_Link_Graph) for the cross-regional pattern weight — those survive. The proposal re-homes the orphaned FIELD to the spine v2.4 brief (which SSBL §9.7 Protocol 7 [L668] already defines as the per-chapter L3-check consumer) and keeps the pattern-weight in the surviving registries. That is data-model reconstruction of an existing CVD structure, not new canon. Each seed also holds the §18.3 line: none declares the thesis, none hints Grand Sage identity (HL_0044) — the "convergence is the payload" language is meta-analysis in the proposal, never in-game content — and all three preserve the RCM "vril not primary anchor" sentences verbatim, so no chapter is silently promoted to a vril-amplification anchor.

How it builds canon. Instantiates the per-chapter L3 carrier convention the runbook can encode once, then reuse for all 77 chapters — closing CANON_AUDIT Known-issue #4's "per-chapter structured anchors largely unpopulated" gap at its root. Stands up the first post-T2 L3 storage precedent (spine brief = field; three §18.5 registries = weight), adds VS_CH34_001 as the first sanctioned expansion-clause append under T1_Vril §5.5 [L172] (the nine canonical anchors are untouched; Canary Islands is genuinely absent from them), and fills three legitimately-empty time_fluctuation_range cells. It also cleanly hands the Ch 34/35 cross-cultural links to the already-running link-graph stream (LINK_0078 Tifinagh-Guanche is ALREADY landed), so the seeds do not duplicate that work.

Why this option. A beats C because CANON_AUDIT Known-issue #4 explicitly rules the foundation-readiness gaps should be closed BEFORE/early in the cascade, not deferred into it; C also leaves HL_0047 with no storage home in the interim. A beats B on the substance of the seeds (identical there), and I absorb B's one correct point: time_fluctuation_range is unschematized and its 40 existing values are era/place SETTING windows, not L3-mechanic strings — so the honest framing is that the spine v2.4 brief is THE per-chapter L3 carrier, the three §18.5 registries carry pattern weight, and time_fluctuation_range carries the time-fluctuation environmental-routing datum (defensible because T1_Vril §11.1 [L319] makes time fluctuation a vril-instability phenomenon) but is NOT elevated to "the" L3 field. That refinement should be written into the runbook so a future reader does not mistake the column for a general L3 store.

Three-layer fit.

• Surface (L1): L1 stays self-contained and untouched: the Ch 34-36 action-adventure crossing (archipelagic landfall, open-ocean traversal, Castilian-conquest and slave-trade-era antagonist staging) reads complete without any seed. Every seed operates a mechanic the surface player already touches — Risco Caído solar-alignment site vril routing, the Kriolu inscription-ability at formation-contact, Trade 10 wayfinding field-mechanics, the ambient Atlantic-vril-field gradient and time-fluctuation regime — so the substrate is operational, never optional (CVD §18.2 L1138).
• Mid (L2): L2 investigatable content is already dense in the RCM for all three chapters (Guanche archaeoastronomy + Libyco-Berber inscription + xaxo mummification; Kriolu creole-formation + Morna; volta-do-mar + rutter navigation corpus). The seeds add the L3 layer without disturbing L2; the Cross_Cultural_Link_Graph rows (Tifinagh-Guanche LINK_0078 landed; Kriolu creole-chain pending) are the L2/L3 hinge the depth player follows.
• Deep (L3): L3 canonical-truth surfaces strictly through Pattern, never spoken: the Atlantic-vril-field intensification-gradient building toward Ch 38, the Guanche flood-memory joining the Thread 4 cross-cultural catalog, the Kriolu convergence chain forward to Ch 42/43/50. Critically, no seed names the convergence as thesis and none hints Grand Sage identity (HL_0043/HL_0044 held) — the payload is left for the player to notice, exactly as §18.5 L1180 requires.

Test (consistency). Before applying, run the standing fresh-context adversarial critic gate on the three RCM appends against: (1) HL_0043/HL_0044 — confirm no in-fiction thesis declaration or pre-Ch-76 Grand Sage/collective-consciousness hint in the paste-ready prose; (2) the RCM 'Vril amplification/substrate not Ch 34/35/36 primary anchor' sentences — confirm each is preserved verbatim and the seed sits at substrate-perception/environmental intensity, not amplification; (3) CVD §17.14/HL_0063 formatting — plain prose, single-level dashes, no code blocks, no emphasis asterisks, affirmative framing; (4) Threads_22 line 145 vs Chapter_Index T02:A — confirm the Ch 35 anomaly-chart material is framed as pattern-seeding environmental routing, NOT a Thread 2 anchor, so it stays valid under either resolution of that open discrepancy. Then verify VS_CH34_001 leaves the nine canonical anchors (Serapeum, Puma Punku, Antikythera, Great Pyramid, Baalbek, Göbekli Tepe, Sacsayhuamán, Nan Madol, Newgrange) unmodified. _Expected:_ Critic returns no HL_0043/0044 violation, all three 'not primary anchor' sentences intact, §17.14-clean, Ch 35 chart = pattern-seeding not Thread-2 anchor, nine vril anchors untouched. On pass, apply the RCM appends + the three time_fluctuation_range cells + (on Josh's row-spec confirm) VS_CH34_001.

Adversarial review — HOLDS. Refuted on all six axes; nothing material broke.

(1) INVENTED CANON — attacked hardest, held. I independently verified every load-bearing anchor by reading the source (not trusting the proposal's citations): CVD §18.4 L1162 ("Every chapter has at least one Layer 3 mechanical event… T2 Chapter Design briefs carry the Layer 3 anchor field") and §18.5 L1178 (the three registries T0_Inscription_Spine / T0_Vril_Site_Registry / T0_Cross_Cultural_Link_Graph) — both verbatim as quoted. RCM Ch 34 verbatim names "Risco Caído… and Cuatro Puertas" and "Component 11 Frame A archaeoastronomical-precision substrate," so the site mechanic and component anchor are canon-named, NOT assigned by the proposal. Threads_22 L700 verbatim contains the Kriolu inscription-ability mechanic. Trades §3.10/§7.2 + CVD §14.6 L795 ("Trades are the most direct Layer 3 mechanical routing") verify the Ch 36 trade-mechanic. Vril §11.1 L319 verbatim: "Time fluctuation is canonically a vril-instability phenomenon." Vril §5.5 L172 verbatim sanctions site-registry expansion beyond the nine anchors. The genuinely-new data (VS_CH34_001 row specifics, time_fluctuation windows, Ch 35/36 component tags) is correctly quarantined into [DECISION NEEDED - JOSH] sub-flags, not asserted as canon.

(2) AUTHORITY HIERARCHY — clean. CVD §18.4/§18.5 is treated as controlling; SSBL (T99 §9.7) is cited only as the downstream consumer, never as binding authority. No T-doc elevated over CVD.

(3) HARD-LINE / §18.3 SMOOTHING — none found. HL_0047 text in the hard-lines CSV matches the proposal's verbatim quote exactly (cross_ref HL_0043,HL_0070 confirmed). I read all three paste-ready RCM appends against HL_0043 (thesis-never-declared) and HL_0044 (no pre-Ch-76 Grand Sage hint, verified verbatim in CSV): every append stays at substrate-perception / environmental-intensity language; none declares the thesis or hints Grand Sage identity. The "convergence is the payload" phrasing appears ONLY in grounding/meta-analysis prose (proposal L31) and the three_layer block — never in in-game paste content. Each append preserves the RCM "vril… not primary anchor" sentence verbatim (RCM Ch 34 confirmed).

(4) STRONGER OPTION — the recommendation already synthesized it. It adopts A but absorbs Option B's one valid point: I independently confirmed via grep that time_fluctuation_range has NO Schema-Dictionary definition (unschematized) and its 14 populated cells are era/place windows (CH_09 "Italian-fascist-era 1936-1941", CH_12 "Atlantic slave trade 1500-1888 CE"), so demoting it from "THE L3 field" to a time-fluctuation datum is correct. CH_34/35/36 cells confirmed empty (data-completeness, not overwrite).

(5) THREE-LAYER — holds. L1 self-contained (each seed operates a mechanic the surface player already touches), L2 already dense in RCM, L3 through Pattern only. Consistent with §18.2 L1138-1142.

(6) DISPOSITION — appropriately conservative, not aggressive. await_josh_confirm correctly gates all naming/numerical/data-model-rehome calls; Vril_Site_Registry confirmed at 30 rows with canonical_status enum {CASCADE_CONFIRMED, LOCKED} and no existing Canary/Risco row, so the sub-flag that neither enum value fits a pre-cascade append is accurate. The Threads_22 L145 ("Ch 34-37… Thread 2 no anchor") vs Chapter_Index T02:A discrepancy is real (I confirmed CH_34-37 all carry T02:A) and correctly routed to the Thread_Grid repair work order as non-blocking.

Stronger option the adversary surfaced: none — the recommendation already synthesized Option A with Option B's one correct refinement (demoting time_fluctuation_range from the general L3 field to a time-fluctuation datum). Option C (defer to cascade) is correctly rejected because CLAUDE.md Known-issue #4 explicitly rules foundation-readiness gaps close BEFORE/early in the cascade, and deferral leaves HL_0047 with no storage home in the interim.

Corrections to fold in:

• Minor, non-blocking: the recommendation's disposition should NOT elevate time_fluctuation_range to any L3-carrier status even nominally — grep confirms the column is absent from T0_Schema_Dictionary v1.3. Best practice before populating CH_34/35/36 cells is to add a one-line definition of the column to the Schema Dictionary in the same pass (recommendation already flags it as unschematized but stops short of ordering the schema patch). This is a completeness note, not a flaw in the recommendation.
• Confirm at apply-time that the RCM append insertion points are recomputed against the live file — the proposal cites lines ~4023/4079/4134 but the Ch 34 section header sits at 4019 and Ch 35 at ~4076 in the current file; append to the END of each chapter's Verified-canon subsection by content match, not by hardcoded line number.

Apply actions. ON JOSH CONFIRM (await_josh_confirm — all four sub-decisions are canon/naming/numerical or an authority-level data-model re-home). Sequence, each edit gated by the standing fresh-context adversarial critic before write:

(1) STORAGE CONVENTION -> encode in docs/CASCADE_RUNBOOK.md: add a short 'Per-chapter Layer-3 anchor (HL_0047) carrier' subsection stating — the spine v2.4 per-chapter brief carries the Layer-3 anchor FIELD (re-homed from the retired T2 briefs per CVD §18.4); the three §18.5 registries (T0_Inscription_Spine, T0_Vril_Site_Registry, T0_Cross_Cultural_Link_Graph) carry cross-regional pattern weight; time_fluctuation_range carries the time-fluctuation environmental-routing datum only and is NOT the general L3 field (it is unschematized). Add the same field to the SSBL §9.6 spine-brief column set if not present.

(2) RCM APPENDS -> file _source/01_Tier_1_Foundation/T1_Research_Corpus_Master [ACTIVE v1.1.4].md: append the three paste-ready 'Layer 3 substrate surfacing at Ch 34/35/36 per HL_0047...' paragraphs (proposal sections CH_34/CH_35/CH_36) to the END of each chapter's '### Verified canon' subsection at lines ~4023 / ~4079 / ~4134. §17.14-clean prose; each preserves the existing 'vril... not primary anchor' sentence verbatim (do not delete/alter it).

(3) VS_CH34_001 -> append one row to registries/T0_Vril_Site_Registry [ACTIVE v0.1]/Sheet1.csv using Josh's confirmed specifics from open_subflag #2 (site_class, vril_function, canonical_status enum value, mastery_tier_required, one-vs-two rows). Do NOT touch the nine canonical-anchor rows. Re-run the harness fidelity gate afterward and record the intended new row (this registry is currently at 30 data rows; expected delta = +1 or +2).

(4) time_fluctuation_range -> set CH_34/CH_35/CH_36 (recommend also CH_33/CH_37) cells in registries/T0_Chapter_Index [ACTIVE v1.2]/T0_Chapter_Index DRAFT v1.1.csv col 8 to Josh's confirmed windows (open_subflag #3). Cells are currently empty; this is data-completeness, not an overwrite.

(5) Cross_Cultural_Link_Graph -> NO action here for Ch 34 Tifinagh-Guanche (LINK_0078 already landed). Hand the remaining Ch 34 (xaxo parallel-mummification, flood-memory Thread 4) and Ch 35 (Kriolu creole-chain to Ch 42/43/50) rows to the existing link-graph population work order; do not duplicate them in this patch.

(6) Component/thesis tags (open_subflag #4) and the Threads_22-vs-Chapter_Index T02 discrepancy (open_subflag #5) are recorded, NOT applied here — the former awaits Josh's component call, the latter feeds docs/proposals/THREADGRID_REPAIR.md.

Commit as a single 'HL_0047 L3 seeds Ch 34-36 + storage convention' change once the critic passes and Josh confirms the row/window/component specifics.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Storage-convention ratification (HL_0047 carrier): confirm spine v2.4 brief = the standing per-chapter L3 anchor FIELD; three §18.5 registries = pattern weight; time_fluctuation_range = time-fluctuation environmental-routing datum only (NOT elevated to the general L3 field — it is unschematized and its existing values are era/place setting windows). This is data-model reconstruction of an existing CVD §18.4/§18.5 structure, but re-homing an orphaned CVD field is an authority call Josh should ratify so the runbook can encode it once.
• [DECISION NEEDED - JOSH] VS_CH34_001 row specifics (Josh-exclusive per cardinal do-not-invent rule): (a) one combined Risco Caído+Cuatro Puertas row vs two rows; (b) site_class = vortex_anchor (Gebel Barkal precedent, VS_CH14_002) vs tomb (Newgrange precedent VS_CH24_001 — but Risco Caído is an almogarén sanctuary, not a tomb); (c) vril_function = substrate_perception (deliberately avoids amplification per the RCM not-primary-anchor sentence) vs harnessing|substrate_perception like Newgrange; (d) canonical_status value for a fresh pre-cascade append — existing enum shows only LOCKED and CASCADE_CONFIRMED, neither fits (recommend a PROPOSED/DRAFT status added to the enum); (e) thread_anchors NULL (no site-to-thread mapping is spelled in canon); (f) cvd_anchor = Component 11 (canon-named in RCM Ch 34 text) with mastery_tier_required NULL (Newgrange carries 6.0 — decide whether Risco Caído substrate-perception implies a tier).
• [DECISION NEEDED - JOSH] time_fluctuation_range window choice per chapter (canon call — what the involuntary fluctuation SHOWS): Ch 34 = Castilian conquest 1402-1496 and/or pre-Achamán cataclysm-memory (both documented substrate); Ch 35 = Ribeira Grande 1462 / 15th-16th c. slave-trade operational peak (moral-weight window, precedent at CH_12); Ch 36 = Henry-the-Navigator era / post-1488 maritime regime. Recommend patching CH_33 and CH_37 in the same pass for arc coherence (Ch 37 candidate: Threads_22 L296 'Atlantic sailor's partial knowledge of pre-flood landscape continuity anchor').
• [DECISION NEEDED - JOSH] Component/thesis anchor assignments (Josh-exclusive): Ch 34 = Component 11 already canon-named in RCM (no decision, listed for completeness); Ch 36 = Component 7 proposed via CVD §14 Layer-3-anchor line ('metaphysical real — trade mastery unlocks substrate perception'), confirm; Ch 35 = NO component named anywhere in canon for the Kriolu-formation convergence — assign one (Component 3 cultural-convergence or Component 7 both plausible) or rule the Thread 12/22 anchors carry the L3 weight without a component tag.
• [DECISION NEEDED - JOSH] Cross-doc discrepancy surfaced during grounding (feeds the Thread_Grid repair work order, NOT blocking these seeds): Threads_22 L145 rules 'Ch 34-37 Atlantic-crossing Thread 2 no anchor per open-ocean chapter-type', but Chapter_Index thread_anchors_complete carries T02:A for CH_34-37. One is wrong. The Ch 35 anomaly-chart seed is framed as pattern-seeding, not a Thread 2 anchor, so it stays valid either way.

---

12. Folklore_Entity_Registry: retarget to Creature_Roster (A) vs resurrect (B)

• Key: folklore_entity_registry_disposition Type: grounded Disposition: await_josh_confirm

Options

• Option A — Retarget to Creature_Roster (honor the recorded absorption) — Treat the registry as retired per the three corroborating records. Retarget BPC §8's stale live-target references (L1923/L1944/L2031 + the §8.11 catalog lists L2157/L2170/L1894) into the existing Creature_Roster expansion stream (BPC L1929 already specs that stream); fold §8.6's per-entry adds into the Creature_Roster adds (only primary_source_citation_text/corpus_query_provenance are net-new there per L2031 vs L2043); repoint Chapter_Index's 79 [POPULATE→T0_Folklore_Entity_Registry] cells at folklore-class creature_ids in Creature_Roster (species_class=folklore, filtered by familiar_slot_eligible/folklore_behavior_substrate). No new registry created. Reconcile the folklore_attribution vs live folklore_attribution_source column name as a paired data-model fix.
• Canon support: [SRC: SCHEMA L926] 'absorbed Folklore_Entity_Registry function in Phase 5M-prep triage per userMemories'; [SRC: SCHEMA L943] folklore_attribution column '(text per Phase 5M-prep triage absorbing retired Folklore_Entity_Registry)'; [SRC: BPC §2.11 L496] 'archived in Phase 5M-prep triage with function absorbed into T0_Creature_Roster folklore_attribution column'; [SRC: BPC §2.2 L322] 'the registry absorbed Folklore_Entity_Registry function in Phase 5M-prep triage'; [SRC: BPC §2.6 L398] 'the folklore_attribution column already specified at Phase 5M-prep triage absorbing the retired Folklore_Entity_Registry'; [SRC: registries/ ls] no folklore registry exists in 29 dirs or _source/00; [SRC: Creature_Roster/Sheet1.csv header] live columns species_class, familiar_slot_eligible, folklore_behavior_substrate, folklore_attribution_source present; [SRC: BPC L1929] a live 'T0_Creature_Roster expansion stream — folklore-entity creature substrate' already exists to receive the fold. No CVD anchor names a Folklore_Entity_Registry: [SRC: CVD grep, zero hits] registry composition is data-model, not a hard line.
• Option B — Resurrect the registry (make BPC §8 literally true) — Author a new T0_Folklore_Entity_Registry with the §8.2/§8.6 schema, populate it from Phase 3 extraction, and leave Chapter_Index pointed at it. Reverses the Phase 5M-prep absorption; creates a second home for folklore-creature substrate alongside Creature_Roster's absorbed columns.
• Canon support: Support is only the STALE half of BPC §8: [SRC: BPC §8.2 L1923] 'T0_Folklore_Entity_Registry stream — every named folkloric supernatural entity'; [SRC: BPC §8.3 L1944] 'streams targeting T0_Folklore_Entity_Registry read T0_Folklore_Entity_Registry row schema'; [SRC: BPC §8.6 L2031] 'minimum-viable adds at Phase 3 stream population'. But these are directly contradicted by the four retirement records above; and the schema they reference ('row schema') exists nowhere — [SRC: SCHEMA grep] no §-entry for the registry, only the Creature_Roster absorption note. Resurrecting means inventing that schema and splitting folklore substrate across two registries against the recorded single-home decision.

Recommendation — Option A — Retarget to Creature_Roster (honor the recorded absorption)

How it fits the CVD. The CVD never names a Folklore_Entity_Registry — registry roster composition is a T0 data-model concern subordinate to the T1 docs, not a CVD hard line [SRC: CVD grep for 'folklore.entity|creature.roster', zero hits]. What the CVD does bind is how folklore creatures are depicted, and that binding is registry-agnostic: [SRC: BPC §2.3 L340] 'CVD §6 cultural depiction standard binds folklore creature representation to authentic per-region tradition substrate' and [SRC: BPC §8.3 L1944] invokes 'CVD §17.1 cultural hard lines for community-protection discipline'. Those §6/§17.1 care standards attach to the folklore-class rows wherever they live, so Option A preserves every CVD obligation while honoring the recorded design decision. Option A also serves Pillar 5 convergence — cross-cultural folklore links live in the separate Cross_Cultural_Link_Graph [SRC: BPC §8.2 L1924], so a dedicated folklore-entity registry is not needed to carry the convergence thesis [SRC: CVD L9 folklore-is-real convergence substrate].

How it builds canon. Retargeting instantiates the already-decided data model: it makes the live registry set self-consistent (one home for folklore-creature substrate, matching the three retirement records), unblocks BPC §8's Phase-3 extraction by pointing it at a schema that actually exists (Creature_Roster's 25 live columns incl. folklore_behavior_substrate/folklore_attribution_source), and clears the 79 dangling Chapter_Index FK placeholders that currently resolve to nothing. It also forces the paired grounded fix — the folklore_attribution (Schema/BPC) vs folklore_attribution_source (live header) name mismatch — which the adapter §2 Blockers already flags, tightening FK integrity ahead of the harness coverage gate. It builds forward toward the Phase-6 Creature_Roster population without standing up a parallel registry that the cascade would have to keep in sync.

Why this option. A honors the recorded canon; B reverses it. Four independent records (Schema L926, Schema L943, BPC §2.2/§2.6/§2.11) say the registry is retired and its function absorbed; only the stale half of one contract (BPC §8) still treats it as live, and that half references a 'row schema' that exists nowhere [SRC: SCHEMA grep]. The authority hierarchy resolves this cleanly: the Schema Dictionary is the schema authority and it records absorption, so BPC §8's live-target language is the error, not the retirement note. Ground truth confirms A is already the de-facto state — no registry exists, the absorbing columns are live — so A is a documentation/re-pointing reconciliation with zero registry-invention, while B would invent a schema, split folklore substrate across two homes, and re-open a closed triage decision for no canon gain.

Three-layer fit.

• Surface (L1): L1 SURFACE: folklore creatures are fully player-facing — tameable populations and folklore-combat encounters per region [SRC: BPC §2.2 L322 'folklore creatures per regional tradition... and tameable populations'; BPC §6.3 L1333 'Folklore combat quests per T0_Creature_Roster per region']. The player fights/bonds them mechanically; which registry table backs them is invisible at the surface. A vs B is player-identical at L1 provided the rows exist — and Creature_Roster is the live table that already carries species_class, familiar_slot_eligible, spawn_rate, and bonding_quest_id, so A delivers the surface layer with no gap.
• Mid (L2): L2 MID/DEPTH: the investigatable substrate — folklore_behavior_substrate (canonical mythological behavior per regional tradition where Layer 2 depth diverges from documented zoology) and folklore_attribution_source — is exactly the absorbed folklore data, and it lives on Creature_Roster today [SRC: Creature_Roster header; BPC §7.3 L1638 'folklore_behavior_substrate text capturing canonical mythological behavior... where Layer 2 depth substrate diverges from documented zoology']. A keeps this depth layer intact and single-homed; B would fragment it across two registries, risking a documented-worldview seam the L2 player could hit.
• Deep (L3): L3 DEEP: the convergence thesis (folklore-is-real across disconnected cultures, never declared) is carried by the Cross_Cultural_Link_Graph and the Grimoire links [SRC: BPC §8.2 L1924 cross-cultural link stream; CVD L9/L43 convergence-as-substrate], not by whether folklore entities have their own registry. A is L3-neutral: it changes no canonical truth, only where the row sits. No pre-Ch-76 surfacing is touched.

Test (consistency). Confirm the retirement is ground-truth and the absorbing home is real, then confirm A leaves no dangling FK: (1) ls registries/ and ls _source/00_Tier_0_Master_Indices/ show NO Folklore_Entity_Registry (done: 29 dirs, zero); (2) head -1 'registries/T0_Creature_Roster [DRAFT v0.1]/Sheet1.csv' shows folklore_attribution_source + folklore_behavior_substrate + species_class + familiar_slot_eligible (done); (3) three retirement records read verbatim at SCHEMA L926/L943, BPC §2.11 L496, §2.2 L322, §2.6 L398 (done); (4) Chapter_Index has 79 [POPULATE→T0_Folklore_Entity_Registry] cells that A repoints (done: grep count 79). After apply, re-grep Chapter_Index for 'T0_Folklore_Entity_Registry' → expect 0 hits; grep BPC §8 for the live-target language → expect it retargeted to T0_Creature_Roster. _Expected:_ Pre-apply: registry absent; Creature_Roster carries the absorbing columns; 4 retirement records present; 79 stale Chapter_Index placeholders + BPC §8 live-target refs present. Post-apply (on confirm): 0 references to T0_Folklore_Entity_Registry outside historical/retirement notes; Chapter_Index cells resolve to folklore-class creature_ids; harness FK/coverage gate passes on the repointed column.

Adversarial review — HOLDS. I independently re-read every cited anchor rather than trusting the dossier. The substantive case for Option A is SOUND and, contrary to my first grep (which missed markdown backslash-escaping), the anchors overwhelmingly check out.

VERIFIED verbatim:

• Retirement records (Option A support): SCHEMA L926 "absorbed Folklore_Entity_Registry function in Phase 5M-prep triage per userMemories"; SCHEMA L943 "folklore_attribution (text per Phase 5M-prep triage absorbing retired Folklore_Entity_Registry)"; BPC §2.2 L322 "the registry absorbed Folklore_Entity_Registry function in Phase 5M-prep triage"; BPC §2.6 L398 "already specified at Phase 5M-prep triage absorbing the retired Folklore_Entity_Registry"; BPC §2.11 L496 "archived in Phase 5M-prep triage with function absorbed into T0_Creature_Roster folklore_attribution column." All five present and accurate.
• Stale §8 live-target refs: L1923 (stream), L1944 (row schema x2), L2031 (minimum-viable adds), catalog lists L1894/L2157/L2170, and the receiving L1929 Creature_Roster expansion stream — all present at the EXACT line numbers the dossier cites. (Note: the older docs/PIPELINE_ADAPTER.md L179 cites different numbers L1943/1964/2051 for the same content; the DOSSIER is correct and the adapter is stale — worth reconciling.)
• Ground truth: ls confirms NO Folklore_Entity_Registry in registries/ or _source/00 (verified both). Creature_Roster header carries species_class, familiar_slot_eligible, folklore_behavior_substrate, folklore_attribution_source and has 0 data rows (header only). Chapter_Index has EXACTLY 79 [POPULATE→T0_Folklore_Entity_Registry] cells, all in the folklore_entities_in_chapter column. Every ground-truth claim in the test block holds.

Authority hierarchy correctly applied: SCHEMA L11 makes the Schema Dictionary the schema-authority layer, and it records absorption; the T1 BPC §8's live-target language is the internal contradiction, not the retirement note. Option B would invent a 'row schema' that exists nowhere and split folklore substrate across two homes against the recorded single-home decision — correctly rejected. Disposition await_josh_confirm is appropriately conservative (registry composition is data-model, Josh's call per the cardinal rule; no apply_now overreach). Three-layer analysis holds: L1/L2/L3 are player-identical under A provided rows exist, and Creature_Roster carries the L2 substrate columns today.

THE ONE MATERIAL DEFECT (a false grounded claim): The dossier asserts repeatedly, as verified fact, '[SRC: CVD grep, zero hits]' and 'The CVD never names a Folklore_Entity_Registry.' This is FALSE. CVD Pillar 5 at L198 explicitly reads: 'Cross-ref: T0_Cross_Cultural_Link_Graph; T0_Folklore_Entity_Registry; T1_Threads_22_Master_References.' The CVD DOES name the registry. This is a false absence-claim presented as grounded — the cardinal-rule failure mode in mirror image.

Does it flip the verdict? No. I verified the CVD L198 mention is a non-binding cross-ref POINTER, not a roster mandate: Pillar 5's binding content is the link mechanic (executable test: 'Link-action without scholarly basis = pillar violation'), and §17.12 Cardinal numerical canon (which DOES enumerate the count hard lines — 72 weapons, 33 species, 22 threads/slots) contains NO Folklore_Entity_Registry entry; no §17 hard line names it either. Registry composition is explicitly delegated to T1_AaS/Schema-Dictionary (SCHEMA L11). So the CVD L198 reference is the same class of stale pointer as BPC §8's live-target language, and Option A preserves the Pillar 5 mechanic fully (folklore entities persist as Creature_Roster rows; cross-cultural links persist in the live Cross_Cultural_Link_Graph). Option A honors every CVD obligation. But the dossier's apply_actions are INCOMPLETE — the retarget list omits CVD L198, which would be left dangling at a retired registry name post-apply — and the false 'zero hits' grounding must be struck. Because the leftover pointer lives in the CVD (constitutional top of the hierarchy), that one edit should be a flagged grounded-pointer fix Josh confirms, not folded silently.

Net: Option A is canon-faithful and correct; the recommendation survives, but with the corrections below and a slightly firmer Josh-confirm on the CVD-pointer touch.

Canon-faithfulness flags raised:

• FALSE GROUNDED CLAIM: dossier asserts CVD never names the registry ('CVD grep, zero hits'); CVD Pillar 5 L198 explicitly cross-references T0_Folklore_Entity_Registry. False absence-claim presented as verified grounding.
• INCOMPLETE RETARGET: apply_actions omit CVD Pillar 5 L198, which Option A would leave pointing at a retired registry; must be added as a Josh-confirmed CVD-pointer fix.

Stronger option the adversary surfaced: none — A vs B is the correct framing and A is right. (A refinement, not a missed option: the CVD Pillar 5 pointer names BOTH Cross_Cultural_Link_Graph and Folklore_Entity_Registry together, which slightly strengthens A — the convergence/link data has a live home (Cross_Cultural_Link_Graph, a real DRAFT registry) and the entity substrate has a live home (Creature_Roster), so Pillar 5's two data needs are both satisfied under A without the retired registry.)

Corrections to fold in:

• Strike the false 'CVD grep, zero hits' / 'CVD never names a Folklore_Entity_Registry' assertions from canon_support and cvd_fit. Replace with the accurate finding: CVD names the registry once, at Pillar 5 L198, as a non-binding cross-ref pointer (not in §17 hard lines, not in §17.12 numerical canon). The correct argument for A is NOT 'CVD is silent' but 'CVD's only mention is a stale pointer of the same class as BPC §8's, and Pillar 5's binding content is the link mechanic, which A fully preserves.'
• Add CVD Pillar 5 L198 to the apply_actions retarget list: the 'T0_Folklore_Entity_Registry' token there should be repointed (to T0_Creature_Roster + T0_Cross_Cultural_Link_Graph, matching Pillar 5's dual data home) as a flagged grounded-pointer fix on explicit Josh confirm — CVD edits sit at the top of the hierarchy and get a firmer confirm bar than the T1/T0 edits.
• Reconcile the line-number drift: docs/PIPELINE_ADAPTER.md L179 cites L1943/1964/2051/2177/2190 and Schema lines 928/945 for the same content the dossier cites at L1923/1944/2031/2157/2170 and Schema 926/943. The DOSSIER numbers are the verified-correct ones; update the adapter (or note the offset) so the apply session edits the right lines.
• Post-apply test should additionally grep the CVD for 'Folklore_Entity_Registry' → expect 0 live-pointer hits (only historical/retirement notes), matching the Chapter_Index and BPC §8 post-conditions the dossier already specifies.

Apply actions. On Josh's confirm (registry roster composition is data-model canon — his call per CLAUDE.md cardinal rule), apply as one batched, gate-tested working-branch change; _source edits only on the same grounded-error precedent as the Combat/Integrity fixes:

1. BPC §8 retarget (in-place _source edit on confirm): change L1923 'T0_Folklore_Entity_Registry stream' to fold into the existing 'T0_Creature_Roster expansion stream' (L1929); retarget L1944 to read Creature_Roster row schema (species_class=folklore rows); merge L2031's adds (primary_source_array, primary_source_citation_text, corpus_query_provenance) into the Creature_Roster adds at L2043 (only citation_text/provenance are net-new); drop 'T0_Folklore_Entity_Registry' from the §8.11 catalog lists at L1894, L2157, L2170. Batch with the standing §8 deep-research-supersession amendment if that is also confirmed.

2. Chapter_Index repoint (CSV edit on working branch): replace the 79 [POPULATE→T0_Folklore_Entity_Registry] cells with folklore-class creature_id references into Creature_Roster (species_class=folklore, per-chapter via chapter_first_encounter). Because Creature_Roster is 0 data rows today, the interim value is a retargeted placeholder [POPULATE→T0_Creature_Roster:species_class=folklore] until Phase-6 population; that removes the dead-registry FK immediately.

3. Paired data-model reconciliation (Josh's call, part of adapter §2 Blockers): pick the canonical column name — recommend keeping the live folklore_attribution_source (populated header wins) and correcting SCHEMA L943 + BPC §2.6 folklore_attribution → folklore_attribution_source as grounded fixes; alternative is migrating the live header to folklore_attribution.

4. Gate: run harness/ FK + coverage gate on Chapter_Index's repointed folklore column and a fresh-context critic on the BPC §8 prose edit before any DRAFT→ACTIVE or merge. Do NOT create any Folklore_Entity_Registry.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Column-name reconciliation folklore_attribution (Schema/BPC) vs folklore_attribution_source (live header): data-model canon — recommend folklore_attribution_source (populated layer wins), correct the two doc citations as grounded fixes; Josh confirms which name is canonical.
• [DECISION NEEDED - JOSH] Interim Chapter_Index cell value while Creature_Roster is 0 rows: recommend a retargeted placeholder [POPULATE→T0_Creature_Roster:species_class=folklore] rather than concrete creature_ids until Phase-6 population — confirm the placeholder convention.

---

13. Verified-count reconciliation: Region 67 (expanding→69) and Questline 0 (target 23) are both faithful populate-states, not migration losses

• Key: verified-counts-region-67v69-questline-23v0 Type: assignment Disposition: await_josh_confirm

Options

• A — Both counts are faithful current populate-states; the 69/23 figures are forward-looking targets. Correct only AAS label drift + remove the duplicate epilogue region row. — Region: 67 data rows is the true migrated state (source xlsx = 67 data rows, live CSV = 67 — byte-faithful) and covers all 79 nodes with zero missing chapters (Ch 2-69 each rowed, Flores Ch2-3 and Atacama Ch53-55 grouped, fairy_realm covers Prologue+Ch1+Ch70-77+Epilogue). '69' is an EXPANDING-TO target reached at Phase 5M, not a CVD lock. Questline: 0 data rows is faithful — the frozen source xlsx has exactly 1 title row / 0 data rows, so nothing was lost; it is a genuine empty [DRAFT v0.1] placeholder whose 23-row roster is fully specified in canon and scheduled for Phase 6 populate. Corrective action: (1) fix AAS L83/L95 wording so target-counts read as targets not current-state; (2) delete the duplicate epilogue region row (Epilogue geography already sits inside fairy_realm's chapter range). The 2 remaining rows to reach 69 are Phase 5M grouping/pipeline adds requiring Josh's grouping calls, not a migration bug to patch now.
• Canon support: [SRC: T0_Region_Index [ACTIVE v1.2]/T0_Region_Index DRAFT v1.1.csv rows 2-68 = 67 data rows]; [SRC: _source/00_Tier_0_Master_Indices/T0_Region_Index [ACTIVE v1.2].xlsx tab 'T0_Region_Index DRAFT v1.1' = 67 data rows, matches live]; [SRC: AAS L83 '69 rows expanding through Phase 5M Drive populate per §7.6']; [SRC: T0_Schema_Dictionary §3.3 'active populate at 69 rows covering canonical regions across 77 chapters per multi-chapter region grouping' — a grouping model, and expands 'to populate scope at Phase 5M']; [SRC: CVD §17.12 L1076-1085 — cardinal numerical canon lists chapters/weapons/species/familiars/chronology/words/sources but NOT a region count → no 69-region hard lock]; [SRC: _source/.../T0_Questline_Index [DRAFT v0.1].xlsx tab Sheet1 = 1 title row, 0 data rows]; [SRC: T0_Questline_Index [DRAFT v0.1]/Sheet1.csv = title only]; [SRC: T0_Schema_Dictionary §3.15 'Status pending Phase 6 populate at 23 rows (Main Story plus 10 Trade plus 12 SQ)' + full 12-SQ enumeration]; [SRC: AAS L95 'T0_Questline_Index [DRAFT v0.1] ... 23 rows ... Drive populate at Phase 6' — its own clause dates the 23 to Phase 6]; [SRC: _source/05_Tier_5_Questlines/T5_*.md — 10 Trade + 12 Side + Master Summary prose substrate exists]; [SRC: duplicate: T0_Region_Index row 'epilogue' vs 'fairy_realm' chapters 'Prologue, Ch 1, Ch 70-77, Epilogue' both cover Epilogue]
• B — Treat 67 and 0 as migration losses; reconstruct 2 missing region rows and all 23 questline rows NOW to match the AAS numbers. — Read the AAS '69 rows' and '23 rows' as authoritative current-state, conclude the migration dropped 2 region rows and 23 questline rows, and rebuild them immediately to force the live registries to 69 and 23.
• Canon support: AGAINST: [SRC: source xlsx region = 67 data rows and questline = 0 data rows — the frozen ground truth was already 67 and 0, so nothing was dropped]; [SRC: CLAUDE.md 'Trust content, not labels ... a registry's real state is its rows, not its version tag']; [SRC: AAS caveat in CLAUDE.md 'it lagged ~1 day behind actual work, so trust file content over version labels']; [SRC: AAS L83/L95 themselves both say 'expanding through Phase 5M' / 'Drive populate at Phase 6' — self-dating the numbers as targets]. Reconstructing region rows now would require inventing grouping/naming canon (Josh-exclusive) and populating questlines ahead of their Phase 6 gate.
• C — Split the region_id to REG_NNN and re-key while reconciling counts. — Fold the count reconciliation into a broader region-registry rework: migrate the string-slug region_id to the schema's REG_NNN primary-key format and re-derive the row set, landing on 69 as a byproduct.
• Canon support: PARTIAL: [SRC: T0_Schema_Dictionary §2 L65 'REG_NNN — region ... 3-digit suffices for 69 regions' and §3.3 'region_id (REG_NNN primary key)']; AGAINST bundling: [SRC: WGM L1107 canonical_anchors T0_Region_Index 'haiti_citadelle_settlement_anchor' and L1100 'azores_triple_junction_vril_convergence_anchor' — live prose canon already references slug-style anchors, so a re-key is a cross-doc FK migration in its own right, out of scope for a count question]; [SRC: CLAUDE.md Known-issue #3 catalogs FK/primary-key model items as a separate pass].

Recommendation — A — Both counts are faithful current populate-states; the 69/23 figures are forward-looking targets. Correct only AAS label drift + remove the duplicate epilogue region row.

How it fits the CVD. CVD never hard-locks a region count: §17.12 (read in full, L1076-1085) enumerates cardinal numerical canon — 77 ch + Pro + Epi, 72 weapons, 33 species, 22 familiars, 5,000-year chronology, ~30M words, ~3,900 sources — and pointedly omits any region total, so '69' carries no constitutional weight and 67 violates nothing. [SRC: CVD §17.12 L1076-1085]. CVD §2 (L25) requires 'Every chapter is anchored to a real-world region'; the live 67 rows satisfy that for all 79 nodes with zero gaps, so the region catalog already meets the CVD's actual regional requirement. [SRC: CVD L25]. The questline roster the CVD cares about is §11.1's 12 Side Questlines plus the 10 trades and Main Story — 23 total — which schema §3.15 already enumerates verbatim; leaving the T0 registry empty until its Phase-6 gate does not contradict the CVD, it defers structured instantiation of canon that already exists in prose. [SRC: CVD §11.1 via T0_Schema_Dictionary §3.15]. Option A respects the authority hierarchy: it declines to let a T0-schema/T99-ops label ('69', '23') override the frozen source and the CVD's silence on region count.

How it builds canon. Reading 67 and 0 as honest populate-states, and fixing the AAS labels to say so, hardens the state-of-truth ledger the whole cascade reads from: it stops a phantom 'we lost 2 regions / 23 questlines' from triggering fabricated backfill (the exact invented-canon-at-scale failure CLAUDE.md's cardinal rule guards against). [SRC: CLAUDE.md 'Cardinal rule: do NOT invent canon' + 'Drift ... came from invented canon at scale']. Deleting the duplicate epilogue region row removes a real data artifact (Epilogue is already inside fairy_realm's chapter range and its geography 'varies per ending'), leaving a clean 66-region + fairy_realm set that the Phase 5M expansion can grow deliberately. [SRC: T0_Region_Index rows 'fairy_realm' and 'epilogue']. It also correctly parks the true expansion-to-69 work (which grouped/fairy nodes to split, whether Prologue/Ch1/endgame get their own region rows) as a Josh-directed Phase 5M grouping decision rather than a migration patch — building the pipeline forward on real gates, not manufactured ones. [SRC: AAS L83 'expanding through Phase 5M'; T0_Schema_Dictionary §3.3 'expands to populate scope at Phase 5M']. Questline stays on its native Phase-6 gate with the full 23-row roster pre-verified as reconstructable from schema §3.15 + T5 docs + CVD §11.1, so the eventual populate is derivation, not invention.

Why this option. Option A is the only reading the frozen ground truth actually supports: both the region xlsx (67 data rows) and the questline xlsx (0 data rows) match the live CSVs exactly, so B's 'migration loss' premise is factually false — nothing was dropped. B would force fabrication of grouping/naming canon (Josh-exclusive) and jump the Phase-6 questline gate, both of which violate the do-not-invent cardinal rule. C bundles in a REG_NNN re-key that is a genuine cross-doc FK migration (WGM prose already cites slug anchors) and belongs to the separate FK pass in Known-issue #3, not to a count question — scope discipline. A is the minimal grounded correction: it trusts content over labels (the standing verification discipline), fixes only demonstrable drift (two AAS target-vs-current mislabels) plus one demonstrable duplicate row, and defers every genuinely Josh-exclusive call (region grouping to 69; questline populate) to its proper gate with the roster pre-verified.

Three-layer fit.

• Surface (L1): N/A — player-invisible infrastructure. Region/questline row counts and the epilogue dedup are data-model bookkeeping; the L1 surface player never sees a registry row, only the chapters and quests those rows will eventually drive.
• Mid (L2): N/A — player-invisible infrastructure.
• Deep (L3): N/A — player-invisible infrastructure.

Test (consistency). Verify four anchors before applying: (1) source region xlsx data-row count == live CSV data-row count == 67 (already confirmed: both 67); (2) source questline xlsx data-row count == 0 (confirmed: 1 title row, 0 data rows); (3) live region CSV covers every node Ch 2-69 plus a fairy_realm row spanning Prologue/Ch1/Ch70-77/Epilogue with zero missing chapter (confirmed by row scan); (4) exactly one duplicate-coverage region row exists — epilogue, whose Epilogue is already inside fairy_realm's range. Then confirm the region fidelity gate still reports only the intended 22 ch_cross deltas after the epilogue-row edit is applied to the live CSV (run the harness fidelity gate on the region registry). _Expected:_ Anchors 1-4 all hold as stated; after deleting the duplicate epilogue row from the LIVE csv only, the harness fidelity gate shows the region registry differing from the frozen xlsx by exactly that one intended deletion (documented), no other cells changed; no chapter node loses coverage.

Adversarial review — FLAGGED FOR REVISION. I independently verified every load-bearing anchor. WHAT HELD: (1) Region live CSV = 67 data rows (header + rows 2-68; line 69 blank), and the frozen source xlsx = 67 data rows with last row = epilogue — byte-faithful, so B's "migration lost 2 rows" premise is factually FALSE. Confirmed via openpyxl. (2) Questline source xlsx = exactly 1 title row / 0 data rows; live CSV matches — nothing was lost; it is a genuine empty [DRAFT v0.1] placeholder. Confirmed. (3) CVD §17.12 (read L1076-1085) enumerates cardinal numerical canon — 77ch+Pro+Epi, 72 weapons, 33 species, 22 familiars, 5,000-yr chronology, ~30M words, ~3,900 sources — and does NOT lock a region count, so 67 violates no hard line. Confirmed. (4) Schema §3.15 enumerates the 23-questline roster + Phase-6 gate verbatim; §3.3 says "active populate at 69 rows ... per multi-chapter region grouping ... expands ... at Phase 5M"; AAS L83/L95 self-date the 69/23 as "expanding through Phase 5M" / "Drive populate at Phase 6." So the thrust — treat 69/23 as forward targets, keep 67/0 as faithful populate-states, defer real expansion/populate to native gates, fix only AAS label drift — is SOUND, and the do-not-invent + trust-content-not-labels discipline is correctly applied against B and C. Disposition await_josh_confirm is appropriately conservative.

WHAT BROKE (material, changes the verdict): corrective action #2 — "delete the duplicate epilogue region row" — is REFUTED as unsanctioned canon mutation mislabeled as bookkeeping. It is NOT a duplicate. T1_World_Geography_Master v2.3 (a T1 DOMAIN doc — higher authority than this T0 registry) carries a DISTINCT entry at L934-938: "Epilogue — Three Distinct Conclusions / Location: Human world location varies per ending / Type: Closing scene," listed separately from both the fairy_realm dimensional framing AND the Ch 76-77 waterfall content (L928-932). The live epilogue row's fields are correspondingly distinct: modern_country and coordinates = "varies per ending," time_period = "post-arc closing scene," and last_updated = 2026-05-03 — one day AFTER every other row (2026-05-02) — i.e. it was deliberately authored, not accidentally duplicated. fairy_realm is "N/A (dimensional)"; the Epilogue is a HUMAN-WORLD location — categorically different geography. Deleting the row would (a) INVERT the authority hierarchy (a T0 dedup overriding a WGM entry), (b) move the catalog 67→66, AWAY from the 69 target rather than toward it, and (c) destroy the only carrier of the canonical "varies per ending" Epilogue geography. The recommendation's own §16-pillar-agnostic claim that "Epilogue is already inside fairy_realm's range" reads a chapter-coverage token overlap as a data duplicate — the overlap is in fairy_realm's chapters array (it lists "Epilogue"), NOT evidence the dedicated row is spurious. The AAS-label fix (action #1) and the questline no-op (action #3) are fine; only action #2 breaks.

Canon-faithfulness flags raised:

• Authority-hierarchy inversion: proposed corrective action #2 deletes the epilogue region row on a T0-registry dedup rationale, but T1_World_Geography_Master v2.3 L934-938 carries a DISTINCT 'Epilogue — Three Distinct Conclusions' geographic entry (human-world location varies per ending, closing scene) that OUTRANKS the T0 registry — a T1 domain doc cannot be overridden by a T0 dedup. [SRC: WGM L934-938]
• Unsanctioned canon mutation dressed as bookkeeping (cardinal do-not-invent rule, deletion side): the epilogue row is materially distinct (modern_country/coordinates 'varies per ending', time_period 'post-arc closing scene', last_updated 2026-05-03 — a day after all peers) and is NOT a duplicate of fairy_realm (which is 'N/A (dimensional)', 'pre-game and post-rift narrative'); deleting it destroys the sole carrier of canonical Epilogue-varies-per-ending geography. [SRC: registries/T0_Region_Index .../T0_Region_Index DRAFT v1.1.csv rows 2 (fairy_realm) and 68 (epilogue)]
• Direction-of-fix error: deleting a row moves the live catalog 67→66, i.e. FURTHER from the stated 69 Phase-5M target, contradicting the recommendation's own framing that the remaining rows are 'expansion-to-69' adds. [SRC: T0_Schema_Dictionary §3.3 '69 rows ... per multi-chapter region grouping ... expands ... at Phase 5M'; AAS L83]

Stronger option the adversary surfaced: Option A-minus-deletion: apply ONLY the AAS L83/L95 label-drift correction (targets read as targets) and the questline no-op; RETAIN all 67 region rows including the WGM-backed epilogue row; treat 'reach 69 at Phase 5M' and 'populate 23 questlines at Phase 6' as Josh-directed native-gate work. This preserves the recommendation's sound core (trust content over labels; both counts are faithful populate-states) while removing the one action that inverts the authority hierarchy and destroys distinct canon. It is strictly safer and equally minimal — it fixes exactly the two demonstrable AAS mislabels and touches zero registry rows.

Corrections to fold in:

• Strike corrective action #2 entirely. Do NOT delete the epilogue region row — it is a distinct WGM-backed canonical entry, not a duplicate. [SRC: WGM L934-938]
• Keep the region catalog at 67 rows; the fairy_realm/epilogue chapter-token overlap (fairy_realm's chapters array includes 'Epilogue') is at most a candidate to trim the 'Epilogue' token FROM fairy_realm's chapter list — never to delete the dedicated epilogue row — and even that is a Josh-exclusive grouping call, not an apply-now dedup.
• Retain corrective action #1 (AAS L83/L95 label-drift fix so 69/23 read as forward targets not current-state) and action #3 (no questline rows authored now; stays on its Phase-6 gate) — both are grounded and correct.
• Remove open_subflag #3 ('confirm intent to delete the duplicate epilogue row') and replace it with a flag noting the epilogue row is confirmed-distinct per WGM v2.3 and should be RETAINED; the only live grouping question is how the 2 expansion rows to 69 are added at Phase 5M.

Apply actions. Numerical/data-model in nature, so hold for Josh sign-off; on confirm apply in-repo only: (1) AAS label fix — in _source/01_Tier_1_Foundation/T1_Architecture_and_State [ACTIVE v3.3].md, edit L83 from '69 rows expanding through Phase 5M Drive populate per §7.6' to '67 rows migrated (frozen source populate-state); expands toward the 69-row Phase 5M target per §7.6 pipeline contract column adds — the final 2 rows are Josh-directed Phase 5M region-grouping decisions, not migrated-and-lost data', and edit L95 from 'T0_Questline_Index [DRAFT v0.1] ... 23 rows covering Main Story plus 10 Trade plus 12 Side Questlines. Drive populate at Phase 6.' to '... 0 rows migrated (empty [DRAFT v0.1] placeholder — never populated in Drive source, confirmed against frozen xlsx); 23-row roster (Main Story + 10 Trade + 12 Side) is fully specified in T0_Schema_Dictionary §3.15 + CVD §11.1 + T5 questline docs and populates at Phase 6.' (2) Duplicate-row removal — delete the epilogue region row from the LIVE registries/T0_Region_Index [ACTIVE v1.2]/T0_Region_Index DRAFT v1.1.csv only (NEVER touch _source frozen xlsx), documenting the intentional 1-row divergence in the harness fidelity-gate expected-delta note alongside the 22 ch_cross edits; leave fairy_realm's Epilogue coverage intact. (3) No questline rows are authored now — that stays on its Phase-6 gate. Open sub-flags below capture the two decisions that remain genuinely Josh-exclusive.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] [DECISION NEEDED - JOSH] Region expansion to 69: how should the final 2 rows be reached at Phase 5M? Options include splitting grouped Earth regions (flores_island Ch2-3, atacama Ch53-55) into per-chapter rows, and/or giving Prologue/Ch1/Ch70-77/Epilogue distinct region rows instead of one fairy_realm row. This is grouping/naming canon (Josh-exclusive per the do-not-invent rule); recommend deciding it as part of the Phase 5M region-page pipeline, not now.
• [DECISION NEEDED - JOSH] [DECISION NEEDED - JOSH] Confirm the region_id primary-key model: live data + WGM prose use string slugs (fairy_realm, haiti_citadelle_settlement_anchor) but Schema_Dictionary §2/§3.3 specify REG_NNN. This is Known-issue #3 FK territory; flagging so the eventual 69-row populate uses the intended key format rather than perpetuating the slug/REG_NNN split.
• [DECISION NEEDED - JOSH] Confirm intent to delete the duplicate epilogue region row from the live CSV (grounded dedup, but it changes a registry row count so surfacing for explicit sign-off).

---

14. Puma Punku: correct CVD §3 Component 10 anchor "Ch 53" → "Ch 52"

• Key: puma_punku_component10_ch53_to_ch52 Type: grounded Disposition: apply_now (adversary revises -> await_josh_confirm)

Options

• A — Correct Component 10 to Ch 52 (recommended) — Change the Component 10 canonical chapter anchor 'Ch 53 Bolivia (Puma Punku)' to 'Ch 52 Bolivia (Puma Punku)'. Leaves Component 11 and every downstream source untouched — they already say Ch 52. One-cell prose fix that resolves the CVD-internal contradiction in favor of the value the rest of canon already carries.
• Canon support: [SRC: CVD L114 §3 Component 11] '2. Puma Punku / Tiwanaku (Bolivia, Ch 52) — H-block interlocks'. [SRC: WGM L669-677 Ch 52 — Bolivia / Uyuni] location 'Bolivia. Uyuni Salt Flats, Tiwanaku... Pumapunku precision stonework'. [SRC: WGM L679-687 Ch 53 — Atacama Begins] 'Location: Atacama Desert, Chile'. [SRC: registries/T0_Vril_Site_Registry [ACTIVE v0.1]/Sheet1.csv L22] 'VS_CH52_001,LOCKED,CVD §3.11,...,Puma Punku Tiwanaku,52.0,Bolivia'. [SRC: registries/T0_Chapter_Index [ACTIVE v1.2]/...v1.1.csv L54-55] CH_52='Bolivia / Uyuni... Tiwanaku', CH_53='Atacama Begins'. [SRC: registries/T0_Region_Index [ACTIVE v1.2]/...v1.1.csv L52-53] bolivia_uyuni=Ch 52 (Tiwanaku), atacama=Ch 53-55. Against option: nothing — the only source carrying 'Ch 53' for Puma Punku is [SRC: CVD L107 §3 Component 10] itself.
• B — Change Component 11 (and downstream) to Ch 53 instead — Treat Component 10's 'Ch 53' as the correct value and re-point Component 11, WGM, the Vril Site Registry, the Chapter Index, and the Region Index to Ch 53. Rejected — it would move Puma Punku onto Atacama's chapter and contradict the geography of both chapters.
• Canon support: Supported ONLY by [SRC: CVD L107 §3 Component 10] 'Ch 53 Bolivia (Puma Punku)'. Directly contradicted by every other source: [SRC: WGM L679-687] Ch 53 is Atacama Desert, Chile — not Bolivia and with no Tiwanaku/Pumapunku content; [SRC: WGM L669-677] Ch 52 is the Bolivia/Uyuni/Tiwanaku/Pumapunku chapter; [SRC: CVD L114 Component 11]; [SRC: T0_Vril_Site_Registry L22]; [SRC: T0_Chapter_Index L54]; [SRC: T0_Region_Index L52]. Would relocate a Component 11 anchor-catalog site into the wrong country and wrong chapter.
• C — Flag only, defer edit — Leave both CVD lines as-is and only document the contradiction. Rejected — the discrepancy is already fully verified (CLAUDE.md Known-issue-1 marks it VERIFIED), every downstream source is unanimous, and leaving a self-contradicting CVD in place is a live drift risk for any cascade session that reads Component 10 first.
• Canon support: No positive canon supports leaving the error in place. [SRC: CLAUDE.md Known-issue-1] already records 'Puma Punku / Tiwanaku — VERIFIED... Ch 52 is canonical. Fix = Component 10's Ch 53 → Ch 52. Do NOT touch Component 11.' [SRC: CVD L1206-1209 §19.2] drift-handling protocol calls to surface drift and propose a paste-ready find-and-replace for author review — which is satisfied by applying with a recorded anchor rather than indefinitely deferring.

Recommendation — A — Correct Component 10 to Ch 52 (recommended)

How it fits the CVD. The CVD is the constitutional foundation and must be internally consistent; it currently contradicts itself, with [SRC: CVD L107 Component 10] saying 'Ch 53 Bolivia (Puma Punku)' while [SRC: CVD L114 Component 11] — the authoritative anchor-catalog for engineering-anomaly sites, which explicitly holds 'the canonical anchor catalog' per [SRC: CVD L111] — says 'Puma Punku / Tiwanaku (Bolivia, Ch 52)'. Component 11 is the section the CVD designates as canonical for site-to-chapter anchoring; Component 10's chapter-anchor line is a supporting cross-reference, not the catalog. Correcting the subordinate reference to match the catalog is exactly the CVD-wins-conflicts hierarchy applied within the CVD itself. It also aligns the CVD with WGM (the T1 geography domain owner) and the T0 registries, restoring top-to-bottom hierarchy agreement.

How it builds canon. Removes the last known self-contradiction in the CVD's eleven-component substrate spine, so any Phase 6 spine-cascade session reading Component 10 for Ch 52/53 substrate gets the correct Bolivia=Ch 52 / Atacama=Ch 53 mapping and cannot propagate 'Puma Punku at Ch 53' drift into region pages, Vril-site cross-refs, or chapter design briefs. It locks CVD ↔ WGM ↔ Vril_Site_Registry ↔ Chapter_Index ↔ Region_Index into unanimous agreement on this anchor, closing CLAUDE.md Known-issue-1.

Why this option. Option A is the only choice consistent with the authority hierarchy and the evidence. Six independent sources across all three tiers (CVD Component 11, WGM, and three T0 registries) place Puma Punku/Tiwanaku at Ch 52 and Atacama at Ch 53; the lone dissenter is one clause in Component 10. Option B would corrupt five correct sources to preserve one typo and would put Puma Punku in the wrong country. Option C leaves a verified self-contradiction live in the constitutional document. The fix is a pure grounded-error correction — no chapter reassignment, no naming, no numerical-canon change, no invention — so it clears the do-not-invent bar and matches the grounded-error precedent already applied this program (Combat 'Ch 45-46 Amazon→Southwest'; the three Integrity_Paths §17.5 mis-citations).

Three-layer fit.

• Surface (L1): N/A — this is a canon-consistency correction to an internal cross-reference; no change to what any player experiences. Ch 52 remains the meditative Uyuni/Tiwanaku/Pumapunku beat and Ch 53 remains Atacama regardless.
• Mid (L2): N/A — player-invisible. The L2 investigatable substrate (Grimoire cross-cultural links, Vril-site inscriptions for the Puma Punku anchor) already keys to Ch 52 via T0_Vril_Site_Registry VS_CH52_001; the edit only makes the CVD prose agree with what the substrate layer already encodes.
• Deep (L3): N/A — player-invisible. The Component 11 engineering-anomaly convergence (Puma Punku as one of nine Pattern anchors) is already correctly chapter-anchored; correcting Component 10 does not alter the never-spoken thesis, only removes an authoring-doc inconsistency.

Test (gate). Confirm the target string is unique, apply the edit, then re-run the migration-integrity scan gate and confirm exit 0 (the edit changes 'Ch 53'→'Ch 52' in CVD prose and introduces no corruption signature; the scanner checks structural corruption, not cross-refs, so it stays green). Commands: grep -c 'Ch 53 Bolivia (Puma Punku)' on the CVD (expect 1 before, 0 after); grep -c 'Ch 52 Bolivia (Puma Punku)' (expect 0 before, 1 after); python harness/scan_migration_integrity.py (expect EXIT=0). Consistency anchors to re-verify post-edit: CVD Component 10 now reads Ch 52 and matches Component 11 L114, WGM L669-687, T0_Vril_Site_Registry L22, T0_Chapter_Index L54, T0_Region_Index L52. _Expected:_ Before: exactly one occurrence of 'Ch 53 Bolivia (Puma Punku)' at CVD L107; scan EXIT=0. After: zero occurrences of the old string, one occurrence of 'Ch 52 Bolivia (Puma Punku)', scan EXIT=0, and CVD Component 10 agrees with Component 11 and all downstream sources.

Adversarial review — HOLDS. REFUTATION ATTEMPT — independently re-read every cited anchor; verified direction, hierarchy, hard-line exposure, disposition, and swept for the same error elsewhere.

WHAT HELD (recommendation is correct on all of these):

• CVD L107 Component 10 reads exactly "Ch 53 Bolivia (Puma Punku)" and the string is UNIQUE (grep count 1) — the edit target and the before/after test are accurate. [SRC: CVD L107]
• CVD L114 Component 11 reads "Puma Punku / Tiwanaku (Bolivia, Ch 52)"; CVD L111 explicitly designates Component 11 as "the canonical anchor catalog." So the recommendation's core hierarchy argument (correct the subordinate cross-reference to match the CVD's own designated catalog) is sound and is CVD-wins-conflict applied WITHIN the CVD. [SRC: CVD L111, L114]
• WGM confirms Ch 52 = Bolivia/Uyuni/Tiwanaku/Pumapunku (relief/meditative), Ch 53 = Atacama Desert, Chile (first combat sprint). Ch 53 has zero Bolivia/Puma content. Option B would relocate the site into the wrong country — correctly rejected. [SRC: WGM L669-687]
• Six-plus independent sources place Puma Punku at Ch 52: Vril_Site_Registry L22 VS_CH52_001, Chapter_Index CH_52, Region_Index bolivia_uyuni=Ch 52, SFX_Registry SFX_00008 (VS_CH52_001, 52.0), Thread_Grid THREAD_01/CH_52, and the entire Research_Corpus_Registry Ch 52 book block. All verified. [SRC: registries as listed]
• Direction (edit Component 10, do NOT touch Component 11), disposition apply_now, and the Known-issue-1 delegation are all correct. No invented canon: no chapter reassignment, no naming, no numerical-canon, no hard-line change. Three-layer "N/A player-invisible infrastructure" is honest — the player experience of Ch 52/53 is unchanged either way.

WHAT BROKE (material flaw, does not flip verdict but must be recorded):

• The recommendation's load-bearing factual claim is FALSE. Option A's "Against option: nothing — the only source carrying 'Ch 53' for Puma Punku is Component 10 itself," echoed in why as "the lone dissenter is one clause in Component 10," is contradicted by a full-corpus sweep. TWO T0 registries carry the SAME Puma-Punku-at-Ch-53 error:

(1) registries/T0_Language_Script_Registry [ACTIVE v0.1]/Sheet1.csv line 73, row PUMA_PUNKU_PRECISION — chapter field = 53.0 (note the sibling row TIWANAKU_INSCRIPTION line 72 correctly carries 52.0 with a [53] cross-ref, so the split is real and internal to this same registry).

(2) registries/T0_Inscription_Spine [ACTIVE v0.1]/Sheet1.csv line 68, row INSC_77H_CH53_FRAGMENT — keyed to chapter 53.0, comment "Component 11 anchor Puma Punku H-block interlock."

• Consequence: the CVD-only edit is INCOMPLETE. After it lands, the CVD says Ch 52 while two registries still assert Puma Punku = Ch 53 — the exact CVD↔registry drift the canon_build section claims this fix "closes." The claim that Known-issue-1 is fully closed by the single edit is therefore overstated.

WHY THE VERDICT STILL SURVIVES: The recommended DIRECTION is right, the CVD edit is correct and gate-safe, and doing it is strictly better than not. The flaw is scope/completeness plus one false supporting sentence, not a wrong decision. The CVD edit should proceed as recommended (apply_now, per Known-issue-1). The two registry rows are a genuinely NEW finding: Known-issue-1's literal delegation is "Fix = Component 10's Ch 53 → Ch 52. Do NOT touch Component 11" — it authorizes only the CVD cell, so extending the fix to two ACTIVE registries is beyond the pre-authorized grounded correction and should be flagged for Josh's confirm rather than auto-applied in the same pass. Hence revised_disposition = apply_now for the CVD cell as written, WITH a mandatory added open sub-flag to correct the two registry rows to 52.0 on Josh's confirm (both edits point the same direction as all other canon; PUMA_PUNKU_PRECISION.chapter 53.0→52.0 and INSC_77H_CH53_FRAGMENT→52.0 / rename fragment id, matched to the existing INSC_77H_CH52_FRAGMENT+TIWANAKU_INSCRIPTION 52.0 rows). No canon invented by either fix.

Stronger option the adversary surfaced: A stronger framing the recommender missed: keep A as the disposition but widen the fix to ALL sources carrying the Puma-Punku-at-Ch-53 error, not just the CVD cell — i.e., a bundled grounded correction of CVD L107 PLUS the two registry rows (T0_Language_Script_Registry line 73, T0_Inscription_Spine line 68), with the CVD cell applied now under Known-issue-1 and the two registry rows applied on Josh confirm since Known-issue-1 only literally delegates the CVD cell. This actually closes Known-issue-1 (which the CVD-only fix does not) and eliminates the residual CVD<->registry contradiction the recommendation would otherwise leave live for the spine cascade.

Corrections to fold in:

• Option A / why: the assertion 'the only source carrying Ch 53 for Puma Punku is Component 10 itself' is FALSE. Two T0 registries also carry it: T0_Language_Script_Registry Sheet1.csv line 73 (PUMA_PUNKU_PRECISION, chapter=53.0) and T0_Inscription_Spine Sheet1.csv line 68 (INSC_77H_CH53_FRAGMENT, chapter=53.0).
• canon_build overstates closure: the single CVD edit does NOT lock 'CVD <-> Vril_Site_Registry <-> Chapter_Index <-> Region_Index into unanimous agreement' and does NOT fully close Known-issue-1, because the Language_Script_Registry and Inscription_Spine rows above still assert Ch 53 after the CVD edit lands — a residual CVD<->registry contradiction.
• test scope is too narrow: it re-verifies only Component 11 + WGM + Vril/Chapter/Region registries and asserts the scanner 'checks structural corruption, not cross-refs, so it stays green' — which is exactly why the scan passing does NOT confirm cross-ref consistency; the two stale registry rows will pass the gate while remaining canonically wrong. Add a cross-registry grep for Puma/Tiwanaku chapter values to the post-apply verification.
• apply_actions should add a second flagged (not auto-applied) action: on Josh confirm, correct T0_Language_Script_Registry PUMA_PUNKU_PRECISION chapter 53.0->52.0 and T0_Inscription_Spine INSC_77H_CH53_FRAGMENT chapter 53.0->52.0 (and reconcile the fragment id with the sibling INSC_77H_CH52_FRAGMENT / TIWANAKU_INSCRIPTION 52.0 rows), so all layers agree on Ch 52.

Apply actions. Grounded-error correction, gate-safe, delegated by CLAUDE.md Known-issue-1 ('Fix = Component 10's Ch 53 → Ch 52. Do NOT touch Component 11.'). Apply as a single unique-string Edit:

FILE: _source/01_Tier_1_Foundation/T1_CVD_Creative_Vision_Document [ACTIVE v1.4].md (line 107, within Component 10 'Canonical chapter anchors:').

OLD: Ch 53 Bolivia (Puma Punku)

NEW: Ch 52 Bolivia (Puma Punku)

(The surrounding clause becomes '...Ch 15 Egypt (Trade 9 Inscription ancient home, Serapeum), Ch 52 Bolivia (Puma Punku), Ch 18 Levant (Baalbek), Ch 23 Greece (Antikythera mechanism).')

Do NOT modify Component 11 line 114 or any registry — all already read Ch 52.

POST-APPLY: run python harness/scan_migration_integrity.py and confirm EXIT=0; re-grep to confirm old string count 0 / new string count 1. Then commit with a message noting 'CVD §3 Component 10 Puma Punku anchor Ch 53→Ch 52 (grounded error-fix; matches Component 11 + WGM + T0 Vril/Chapter/Region registries; closes Known-issue-1)'. Since this edits the frozen _source CVD, note it in the ROADMAP §9 ledger / STATUS per the in-repo grounded-fix precedent (mirrors the Combat and Integrity_Paths grounded corrections). No canon invented; no numerical/naming/hard-line change.

---

15. 120 quest-gate provenance: ratify the derivation or author a CVD anchor

• Key: 120_quest_gate_provenance Type: assignment Disposition: await_josh_confirm

Options

• A — Ratify the derivation as already-canon in T1_Trades_Master; add NO CVD number — Rule that 120 is canon because T1_Trades_Master v1.0 ACTIVE §4-§5 already instantiates it as a grounded operationalization of two CVD rules (universal framework §12.1 + every-tier-quest-unlock §17.5). CVD stays silent on the count by design — 120 is a downstream structural fact owned by the domain doc, not a cardinal number CVD must declare. Correct the CLAUDE.md parenthetical from the loose '12 tiers x 10 trades' to the accurate '10 trades x 12 gates', since Trades_Master §4 L99 itself derives it that way and §4.1-4.5 build 12 GATES (10 named tiers + 2 transition-pivot gates), not 12 tiers.
• Canon support: [SRC: T1_Trades_Master §4 L99 'ten trades multiplied by twelve quest gates per trade']; [SRC: Trades §4.1-4.5 L105-134 = 5 Outer + Portal pivot + 3 Inner + Abyss pivot + 2 Third = 12 gates]; [SRC: Trades §5 L140-181 rosters all 120, verified by count: TR1-TR10 each x12, NEO x10, MGS x10]; [SRC: CVD §12.1 L604 universal framework + §17.5 L977 every-tier-quest-unlock — the two rules the roster combines]; [SRC: CVD §14.2 L751-762 = 10 trades; §17.6 L987 names locked]. Number 120 confirmed ABSENT from CVD (full-doc grep).
• B — Author a new CVD §17.12 numerical-canon bullet stating 120 — Add a cardinal-numerical-canon line to CVD §17.12 (L1076-1085) making 120 CVD-level canon alongside 77 chapters / 22 threads / 72 weapons. Elevates the count to constitutional status so no downstream doc can drift it.
• Canon support: [SRC: CVD §17.12 L1076-1085 is where cardinal counts live — 72 weapons, 33 species, 22 slots are all here, so 120 would be topically at home]; but [SRC: CVD §12.1 L610-648 + §17.5 L971-981 deliberately delegate 'specific quest-gate content per tier per capability' to T1_Combat/Trades/Ability at L650] — CVD's own pattern is to NOT enumerate per-capability gate counts.
• C — Flag 120 as unsupported and strip it from CLAUDE.md numerical canon — Treat 120 as an un-anchored invention and remove it from CLAUDE.md's numerical-canon line until Josh authors a CVD anchor. Rejected framing — included for completeness.
• Canon support: [SRC: CVD full-doc grep — 120 absent] would support caution, BUT [SRC: T1_Trades_Master v1.0 ACTIVE §4-§5] fully derives and rosters it from CVD rules, so it is NOT invented — it is a valid domain-doc instantiation. Stripping it would discard authored ACTIVE canon.

Recommendation — A — Ratify the derivation as already-canon in T1_Trades_Master; add NO CVD number

How it fits the CVD. CVD's own architecture explicitly delegates gate counts downstream: §12.1 L650 routes 'specific quest-gate content per tier per capability' to T1_Combat_System_Spec, T1_Trades_Master, T1_Ability_Tree; §17.5 L977 states only the RULE ('every tier requires a quest-unlock'), not a number. The authority hierarchy (CVD > T1 domain > T0) means the trade-gate count is properly owned by T1_Trades_Master, the domain doc for trades. 120 is a faithful product of two CVD hard lines (universal 12-slot framework x 10 locked trades), not free invention. Option A keeps CVD lean and constitutional, exactly as §17.12's existing bullets do for domain-owned structure, while making the number correct and traceable.

How it builds canon. Ratifying resolves the provenance flag WITHOUT touching CVD, and it disentangles this from Known-issue #2: the '12' in '120' is not '12 tiers' (there are 10 named tiers, CVD §12.1 L610-646) — it is '12 GATES per trade' (10 tier-transitions + the Portal-Grade and Babe-of-the-Abyss pivot gates, Trades §4.2/§4.4 L113-127). Correcting CLAUDE.md's parenthetical to '10 trades x 12 gates' makes the numerical-canon line internally consistent and stops the loose '12 tiers' phrasing from re-fusing 120 to the twelve-tier-label mismatch every future reader will trip on. It stands up a clean, citable provenance chain (CVD §12.1+§17.5 -> Trades §4 -> Trades §5 roster of 120 -> T0_Quest_Definition_Registry at Phase 5M) that the Phase 6 cascade and quest-writing pipeline can rely on.

Why this option. A beats B because CVD deliberately does not enumerate per-capability gate counts (§12.1 L650 delegates exactly this); adding 120 to §17.12 would break CVD's constitutional-lean pattern and duplicate a fact the domain doc already owns and could drift against. A beats C because 120 is demonstrably NOT invented — T1_Trades_Master v1.0 ACTIVE §4-§5 derives it from CVD rules and rosters all 120 gates (verified by count), so stripping it would delete authored ACTIVE canon and contradict the do-not-invent discipline's own premise (the number is grounded). A is the only option that both ratifies grounded canon and fixes the one real defect — the mislabeled '12 tiers x 10 trades' phrasing.

Three-layer fit.

• Surface (L1): L1 player: 120 gates are the milestone rhythm across the 77-chapter arc — a quest-unlock at every tier transition per capability, felt as clean progression ceremony (CVD §12.1 L606-608, §16.7 milestone-rhythm L887). Player never counts to 120; they feel the beats.
• Mid (L2): L2 player: recognizes the Golden Dawn / Hermetic grade ladder (Neophyte -> Magus) behind the gates (CVD §12.1 L608), and that trade mastery reveals substrate concepts progressively (§14.3). The 12-gate-per-trade structure is the documented worldview scaffolding behind the milestone shapes.
• Deep (L3): L3: the uniform 12-gate framework across all 10 trades (and all capabilities) is itself the convergence-Pattern in miniature — every documented tradition climbs the same substrate ladder, never declared, only structurally present. The count 120 is infrastructure serving that Pattern, never surfaced.

Test (consistency). Verify the provenance chain holds after the CLAUDE.md wording fix: (1) CVD contains no literal '120' — confirmed by full-doc grep; (2) T1_Trades_Master §4 L99 states '10 trades x 12 gates'; (3) Trades §5 L140-181 rosters exactly 120 distinct gate ids — confirmed: stems TR1-TR10 each appear 12x, NEO x10 and MGS x10 (one per trade); (4) after edit, CLAUDE.md line 18 reads '12-tier mastery (120 quest gates = 10 trades x 12 gates)' and no longer asserts the un-anchored '12 tiers x 10 trades' factorization. No harness gate covers prose provenance; this is a read-and-confirm anchor check. _Expected:_ CVD 120-count absent (correct — delegated); Trades_Master derives and rosters 120; CLAUDE.md parenthetical corrected to the 10x12 factorization; Known-issue #2 (twelve-tier label vs ten named tiers) explicitly noted as SEPARATE from and not the source of the 120 count.

Adversarial review — FLAGGED FOR REVISION. REFUTATION ATTEMPT — the recommendation does NOT survive as written. I independently read CVD §12.1 (L604-650), §17.5 (L971-981), §17.12 (L1076-1085), and T1_Trades_Master §4-§5 (L97-182), and grepped CVD full-doc for "120".

WHAT HELD:

• 120 is genuinely ABSENT from CVD (grep: only "120" hits are Nan Madol Ch 63 and unrelated numbers). Verified.
• T1_Trades_Master v1.0 ACTIVE §5 rosters 12 gate-ids per trade x 10 trades = 120 distinct ids (TR1-TR10 each x12). Verified by reading §5.1-§5.10.
• CVD §12.1 L650 does delegate "specific quest-gate content per tier per capability" to T1_Combat/Trades/Ability. Verified — delegation of gate CONTENT is real.
• Option A beats B and C in the narrow sense argued (don't strip authored ACTIVE canon; don't bloat §17.12). That relative ranking is defensible.

WHAT BROKE (load-bearing failure):

The headline claim — that '120 = 10 trades x 12 gates' is a clean product of CVD rules and INDEPENDENT of Known-issue #2 — is FALSE against the CVD text. CVD §12.1 L610-646 enumerates the twelve-tier framework as grades NUMBERED Tier 1 through Tier 10 (Neophyte=Tier1 ... Magister Templi=Tier9, Magus=Tier10). There is NO 'Portal Grade' tier and NO 'Babe of the Abyss' tier in CVD's list. CVD names Portal Grade only as a CEREMONY inside Tier 5 Philosophus's quest gate (L628); 'Babe of the Abyss' appears NOWHERE in CVD (grep-clean; sole Abyss ref L644 = 'Abyss-crossing' describing the Magister Templi PRESTIGE questline, not a tier/gate).

Trades_Master §4.2/§4.4 (L113-127) INVENTS Portal Grade and Babe-of-the-Abyss as two standalone 'transition-pivot gates' to pad 10 named grades up to 12 gates-per-trade. That is a domain-doc reconciliation of the exact CVD 'twelve-tier label over ten named grades' conflict that Known-issue #2 flags as Josh's standing, UNRESOLVED, preserve-verbatim item. So the '12' in '120' is NOT independent of KI#2 — it is one specific (Trades_Master) resolution of KI#2, and CVD's own §12.1 resolves the same ambiguity a DIFFERENT, incompatible way (12 numbered tiers; Portal Grade as a Philosophus sub-ceremony). The recommendation asserts 'independent of whether the tier LABEL is ever corrected' — that contradicts the CVD text I read.

SECOND BREAK (numerical-canon integrity): Trades §5 counts Gate 12 = Magus into all ten trades' 120. But CVD §17.5 L981 and Trades §4.5 L134 both bind Magus as Tier-C-MMO-endgame ONLY, unavailable to Tier A. So 120 includes 10 gates (one Magus per trade) a solo/Tier-A player can NEVER unlock. The recommendation's L1 three-layer claim that '120 gates are the milestone rhythm across the 77-chapter arc' overstates: the 77-chapter arc is Tier A, whose reachable trade-gate count is 110, not 120.

NET: 120 is not 'invented' in the drift-at-scale sense (Trades_Master derives it transparently), so Option C is correctly rejected. But it is ALSO not the CVD-clean, KI#2-independent fact the recommendation sells. It is a T1-domain reconciliation of an OPEN CVD conflict that lands on a tier-count reading CVD itself does not number that way, and it bundles a Tier-C-only gate into a count framed as the Tier-A arc rhythm. The recommendation should ratify 120 only AFTER Josh rules KI#2, and must not write '120 = 10 trades x 12 gates' into CLAUDE.md as settled cardinal canon while KI#2 is open. Stronger missed option (D / A-prime): ratify 120 as domain-owned but CONDITION it on Josh first resolving KI#2 (12 numbered tiers per CVD, or 10 grades + 2 pivot gates per Trades_Master); annotate CLAUDE.md as 'pending KI#2 ruling' rather than asserting the 10x12 factorization as fact; flag the Tier-A 110 vs A+C 120 wrinkle. Disposition stays await_josh_confirm, but the thing awaiting confirm is the KI#2 tier/gate structure, of which 120 is a downstream consequence — not an independent number the recommendation can pre-settle.

Apply actions. This is a numerical-canon ratification (Josh-exclusive per the cardinal rule), so hold for confirm. On Josh's approval of Option A: (1) No CVD edit. (2) Edit CLAUDE.md line 18: change '12-tier mastery (120 quest gates)' to '12-tier mastery (120 quest gates = 10 trades x 12 gates per trade)'. (3) Edit docs/proposals/ROSTER_PROPOSALS.md item 8 (L417) and docs/archive/LEDGER_BATCH1.md L23: replace the '12 tiers x 10 trades' framing with '10 trades x 12 gates (10 named tiers + Portal + Abyss pivot gates), per T1_Trades_Master §4 L99' and mark the flag RESOLVED-ratified. (4) Optionally add one clarifying sentence to T1_Trades_Master §4 L99 noting the 12 gates = 10 named tiers + 2 transition-pivot gates, to pre-empt the twelve-tier-label confusion (grounded clarification, not new canon). All edits in-repo only; commit with evidence citing CVD §12.1 L650 delegation + Trades §4-§5 count.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Whether 120 should ALSO be elevated to a CVD §17.12 cardinal-numerical bullet (Option B) is Josh's constitutional call — recommendation is NO (CVD delegates gate counts by design), but the elevate/delegate choice is his numerical-canon authority.
• [DECISION NEEDED - JOSH] Known-issue #2 (CVD labels the framework 'twelve-tier' while §12.1 names ten tiers) remains Josh's standing preserve-verbatim item — this recommendation does NOT touch it; it only clarifies that the '12' in the 120-count is 12 GATES (10 tiers + 2 pivots), which is independent of whether the tier LABEL is ever corrected.

---

16. Twelve-tier mastery label vs ten named tiers

• Key: mastery-twelve-tier-label Type: assignment Disposition: await_josh_confirm

Options

• A. Preserve as-is; log the reconciliation note (RECOMMENDED) — Keep 'twelve-tier' verbatim everywhere as the immutable framework name. Do NOT auto-correct to 'ten-tier'. The apparent mismatch is already reconciled in canon: the framework is NAMED for its Golden Dawn provenance (which has more grades than the game exposes), while players access ten enumerated tiers. Add one non-canon documentation line (in the FK/known-issues ledger, not in CVD) recording that '12-tier' = framework-provenance name and '10 accessible tiers' = playable count, so no future agent 'fixes' it.
• Canon support: [SRC: CVD §12.1 L604-648] header 'Twelve tiers' + 'The twelve tiers... :' introduces a list that enumerates exactly Tier 1 Neophyte through Tier 10 Magus. [SRC: ABIL §2 L45] states verbatim 'The ten accessible tiers across the canonical numerical scale' — canon ALREADY distinguishes framework-name from accessible count. [SRC: CVD §12.1 L608] naming 'draws from documented Western mystery tradition (Hermetic / Golden Dawn lineage)'; the Golden Dawn schema historically carries grades beyond the ten exposed (a 0=0 Neophyte + Portal transitional grade in/around First Order, and three grades above the Abyss — Magister Templi, Magus, Ipsissimus — of which canon exposes only two). [SRC: CVD §12.1 L628 + ABIL L53] canon already references the real Golden Dawn 'Portal Grade ceremony' at Tier 5. [SRC: CVD §17.5 L975] hard line locks 'twelve-tier Hermetic mastery framework' as immutable, non-varying. [SRC: CVD §17.12 L1076-1085] cardinal numerical canon does NOT enumerate a tier count, so 'twelve' is not a load-bearing cardinal number that ten would violate.
• B. Surface as intended two-tier expansion (pre-tier + post-Magus capstone) — Treat the gap as deliberate headroom: a Tier 0 pre-initiate (0=0 Neophyte antechamber) below Tier 1 and a Tier 11 Ipsissimus capstone above Magus, to be authored later, making the count literally twelve. Recommend Josh RULE whether these two are reserved design space.
• Canon support: SUPPORTED-BY-PROVENANCE but NOT-STATED: [SRC: CVD §12.1 L608] Golden Dawn lineage does contain a 0=0 grade and an Ipsissimus (10=1) capstone above Magus, so twelve is a natural provenance count. AGAINST: [SRC: CVD §12.1 L646] and [SRC: CVD §17.5 L981] explicitly cap the player path at Magus ('Tier A solo... caps at Magister Templi'; Magus is Tier C MMO endgame) with NO tier named above it — a canonical ceiling, not headroom. [SRC: ABIL L45] 'ten accessible tiers' reads the design intent as ten-accessible, not twelve-eventual. No canon anywhere names Ipsissimus or a Tier 0. This would be canon INVENTION.
• C. Renumber the label to 'ten-tier' to match the enumerated list — Auto-correct every 'twelve-tier' string to 'ten-tier' across CVD, ABIL, VRIL, TRADES so name and count agree.
• Canon support: AGAINST — violates the standing instruction and a hard line. [SRC: CVD §17.5 L975] names 'twelve-tier Hermetic mastery framework' as an immutable hard line; [SRC: Known-issue #2, CLAUDE.md] Josh's standing call is 'preserve verbatim, flag only, never auto-correct.' The string 'twelve-tier' is load-bearing across [SRC: VRIL §9 L252-264, ABIL §2 L41-66, and 8+ CVD occurrences]; a bulk rename is exactly the kind of unflagged canon edit the cardinal rule forbids.

Recommendation — A. Preserve as-is; log the reconciliation note (RECOMMENDED)

How it fits the CVD. Fits the CVD's naming philosophy precisely. [SRC: CVD §12.1 L608] the tier names are chosen to read as 'becoming a practitioner of a tradition' rather than 'leveling up a stat,' and 'Layer 2 depth players recognize the Golden Dawn system.' A Golden-Dawn-literate player expects the framework to be named for its full historical schema even where the game exposes a subset — so 'twelve-tier' as the framework's provenance name with ten accessible grades is more authentic to the source tradition than a sanitized 'ten-tier' would be. It also respects §17.5's immutability hard line [SRC: CVD L975] and the do-not-invent cardinal rule, since Option A changes zero canon text.

How it builds canon. Stands up a durable disambiguation the cascade needs: every downstream doc (Combat Spec, Trades Master, per-weapon/per-trade quest gates) references 'twelve-tier framework' but instantiates ten gates. Recording '12 = framework-provenance name, 10 = accessible/gate count' in the ledger prevents a future spine agent from authoring phantom Tier 11/12 quest gates or mis-counting mastery milestones, and prevents a well-meaning 'consistency' pass from renaming the framework and breaking the §17.5 hard line. It converts a standing ambiguity into a one-line documented invariant without touching canon.

Why this option. A beats C because C violates both the explicit standing instruction ('never auto-correct') and the §17.5 immutability hard line [SRC: CVD L975]. A beats B because B is canon invention: canon deliberately and repeatedly caps the player at Magus [SRC: CVD §12.1 L646, §17.5 L981] with no named grade above and no Tier 0 below — reading the gap as 'reserved headroom' contradicts the explicit ceiling, whereas reading it as 'framework named for its fuller provenance, ten grades exposed' is exactly what [SRC: ABIL L45] already says in canon. A is the only option that both resolves the confusion AND changes no canonical text.

Three-layer fit.

• Surface (L1): L1 players experience ten clean progression milestones (Neophyte through Magus). The label 'twelve-tier' never surfaces mechanically; players see tier names and numerical bands, not the count. No surface impact from preserving the name.
• Mid (L2): L2 depth players recognize the Golden Dawn / Hermetic lineage [SRC: CVD §12.1 L608] and the 'Portal Grade ceremony' [SRC: L628] as real transitional-grade vocabulary. For this player the 'twelve-tier' framing is a FEATURE — it signals the framework is named for the authentic tradition, whose full schema exceeds what the game exposes. Renaming to 'ten-tier' would erase that provenance signal.
• Deep (L3): L3 canonical-truth: the convergent-traditions thesis is served by the mastery ladder BEING a documented real tradition rather than an invented stat scale. The name pointing at a real, larger grade system (with grades the mortal path cannot reach) reinforces that the substrate is real and larger than any single practitioner's reach — the Magus ceiling and the unreachable capstone are thematically coherent with 'the substrate is the air the game breathes, never fully declared.'

Test (consistency). Verify no canon names an 11th/12th playable tier and that the reconciliation holds: (1) confirm CVD §12.1 enumerated list ends at Tier 10 Magus [SRC: CVD L646]; (2) confirm ABIL §2 states 'ten accessible tiers' [SRC: ABIL L45]; (3) confirm §17.5 hard line names 'twelve-tier framework' [SRC: CVD L975]; (4) grep the T1 corpus for 'Tier 11', 'Tier 12', 'Ipsissimus', 'Tier 0' — expect ZERO playable-tier hits (only provenance/omission notes acceptable). If any playable Tier 11/12 reference exists, re-flag to Josh. _Expected:_ Enumerated list ends at Tier 10 Magus; ABIL says 'ten accessible tiers'; §17.5 says 'twelve-tier'; zero canonical references to a playable Tier 11/12, Ipsissimus, or Tier 0. Reconciliation confirmed: label preserved, count documented.

Adversarial review — HOLDS. Refutation attempt on all six attack surfaces; the recommendation held on every one.

ANCHOR VERIFICATION (read independently, not trusted): Every load-bearing anchor checks out verbatim. (1) CVD §12.1 L604 header "Twelve tiers, Hermetic tradition"; L610 "The twelve tiers... :" introduces a list enumerating exactly Tier 1 Neophyte (L614) through Tier 10 Magus (L645) — ten items. CONFIRMED. (2) CVD §12.1 L646 + L648 explicitly cap the player: "Tier A solo completion alone caps at Magister Templi" and Magus is "Tier C MMO endgame only" — a canonical ceiling with NO named grade above Magus. CONFIRMED. (3) ABIL §2 L45 reads verbatim "The ten accessible tiers across the canonical numerical scale." CONFIRMED. (4) CVD §17.5 L975 names "The twelve-tier Hermetic mastery framework" as the immutable hard-line term. CONFIRMED. (5) CVD §17.12 L1076-1085 cardinal numerical canon does NOT enumerate a tier count — so "twelve" is not a cardinal number that "ten" would violate. CONFIRMED (a genuine strength the recommendation correctly identified). (6) Negative grep across all of T1 Foundation for Ipsissimus / playable Tier 11 / Tier 12 / Tier 0 / eleven-twelfth-tier: ZERO real hits (the apparent "Tier 11/12" matches were false positives — "Tier C", "Tier 0 registry", folder-ID substrings). CONFIRMED. "twelve-tier" is genuinely load-bearing: 89 occurrences across 8 T1 files, so a bulk rename (Option C) is exactly the unflagged scaled canon edit the cardinal rule forbids.

ATTACK 1 — invented canon presented as grounded? No. Option A's canon_support cites real-world Golden-Dawn history (0=0 grade, Portal grade, supernal triad) as EXTERNAL provenance, not as claims about Humanity-canon — so it invents no world-canon. The reconciliation A actually rests on is fully canon-internal and already present: ABIL says "ten accessible" while CVD names it "twelve-tier." That is a real existing reconciliation, not an invention. Option B, by contrast, IS canon invention (naming a Tier 0 and an Ipsissimus capstone) and the recommendation correctly recommends AGAINST it, citing the explicit Magus ceiling that contradicts the "reserved headroom" reading. Correct call.

ATTACK 2 — authority-hierarchy error? No. ABIL (T1) "ten accessible" does not CONTRADICT CVD "twelve-tier"; it disambiguates count-vs-name, and CVD would win any true conflict anyway. No T99/draft is treated as binding.

ATTACK 3 — smoothed-over §17/hard-line violation? No. §17.5 locks the STRING "twelve-tier framework," which Option A preserves untouched; §17.12 does not fix a tier count, so no numerical hard line is in tension. Option A is the only option that respects both.

ATTACK 4 — stronger option missed? No. A/B/C span the real space (preserve+document / expand-to-literal-twelve / rename-to-ten). A dominates: it resolves the ambiguity and edits zero canon.

ATTACK 5 — three-layer claim that fails? No. L1 "ten clean milestones, label never surfaces mechanically" is consistent with §12.1 L608 ("Layer 1 surface players experience clean progression milestones"). L2 provenance-recognition is verbatim from L608 ("Layer 2 depth players recognize the Golden Dawn system"). L3 coherence with the Magus ceiling holds. Sound.

ATTACK 6 — disposition too aggressive? No — correctly conservative. await_josh_confirm with NO canon edits honors both the §17.5 hard-line term and Josh's standing Known-issue-#2 instruction ("preserve verbatim, flag only, never auto-correct"). The one proposed change is a non-canon ledger/known-issues line, not a canon-file edit, which is appropriate and does not require the more restrictive treatment. No downgrade warranted.

Verdict: canon-faithful, anchors verified, no material flaw. Survives.

Corrections to fold in:

• Minor framing, not a defect: Option A's canon_support leans on real-world Golden Dawn grade details (a 0=0 grade, three supernal grades) to motivate 'twelve.' These are true of the external tradition but are NOT stated as Humanity-canon and should never be mirrored INTO a canon doc as if the game exposes them — the ledger note should stay strictly count-vs-name ('12 = framework-provenance name, 10 = accessible/gate tiers'), which the recommendation's proposed ledger text already does. Keep the historical-grade rationale in the review record, not in any canon file.
• The proposed ledger line should live in CLAUDE.md Known-issue #2 / the FK-known-issues ledger only (non-canon), never in CVD or ABIL — the recommendation says this, and it must be preserved on apply so no 'twelve-tier' string in a canon file is edited.

Apply actions. Josh-exclusive because it touches a §17.5 hard-line term and a numerical-framework interpretation. On Josh sign-off of Option A: (1) NO edits to any canon file — 'twelve-tier' stays verbatim everywhere. (2) Add one non-canon ledger line to CLAUDE.md Known-issue #2 (or the FK/known-issues ledger), replacing the current open flag with a RESOLVED note: 'Twelve-tier label — RESOLVED (preserve). Framework is NAMED for its Golden Dawn provenance; canon exposes ten accessible/gate tiers (Neophyte..Magus). Ten-accessible language is already canonical at ABIL §2 (\"ten accessible tiers\"). \"Twelve-tier\" is the immutable framework name per CVD §17.5; do NOT rename, do NOT author Tier 11/12 gates. Player path caps at Magus per CVD §12.1/§17.5.' (3) Add the same one-liner to the spine load-manifest / cascade notes so per-chapter agents count ten mastery gates, not twelve. If Josh instead selects B (reserve Tier 0 + Ipsissimus capstone), that is canon invention requiring his authored tier names and gates before any doc edit.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Option B (pre-tier + post-Magus Ipsissimus capstone as reserved design space) would be canon invention — recommended AGAINST on current evidence, but flagged here as the only Josh-exclusive alternative if he intends the count to become literally twelve.

---

17. Boss personal-name methodology + illustrative candidates (naming canon — Josh rules)

• Key: D-BOSS-NAMES Type: assignment Disposition: await_josh_confirm

Options

• A. Substrate-Echo methodology (the §17.7 realm-name discipline generalized) — For each boss whose culture has a documented living linguistic substrate, derive the personal name by ECHOING that substrate — a plausibly-formed name in/adjacent to the documented language and onomastic tradition — never lifting a sacred/registered proper name, never a real deity/ancestor/historical figure's name. Where the antagonist is a Velheim vessel/operative, the name follows the vessel's cultural setting, not a fantasy register. Where canon documents a 'reference-only-with-discipline' taint (§19.2), avoid the tainted proper noun and coin fresh. This is exactly what Josh already ruled into canon for the four threshold realms: Ma'at -> Maatherion, Tiamat -> Tiamarath, Pohjola -> Pohljar (echo, not appropriation).
• Canon support: [SRC: CVD §17.7 L1009-1012 realm names echo a substrate 'Substrate echoes...' and 'Supersedes [legacy]' — the canonical echo-not-appropriate template] [SRC: CVD §17.1 L927-929 'Each culture appears with its own self-naming'; 'Registered and protected cultural property is referenced, not reproduced'] [SRC: CVD §18.3 L1148-1156 show-don't-tell forbids spoon-feeding/appropriative shorthand] [SRC: CVD §19.2 L1010 reference-only-with-discipline for Theosophical/Nazi-occult tainted proper nouns] [SRC: boss_roster_rows.json 'NAMING DISCIPLINE (GRAFT-A, binding)': every TBD-stub carries only its canon title; personal names remain DECISION NEEDED - JOSH]
• B. Descriptive-title-only (assign no personal names at all) — Keep every boss on its canon descriptive title permanently (The Corrupted Hunter, The Corrupted Hierophant of Eleusis...). No personal names anywhere; the title IS the name.
• Canon support: [SRC: boss_roster_rows.json rows — the roster already ships clean descriptive titles for all 75 bosses] [SRC: CVD §18 show-don't-tell favors encounter-over-exposition] — but AGAINST: named vessels already exist in canon (Sayyid Abdallah bin Rashid Ch 8, Meryre Ch 15, Arikankharor Ch 14, Datu Siahaan Ch 5, Raghavendra Ch 7, Oba Ayize Ch 12, Li Si Ch 21), so a blanket no-names rule contradicts established canon [SRC: boss_roster_rows.json Ch 5/7/8/12/14/15/21 boss_name fields]
• C. Author-invent full fantasy names now (Claude assigns) — Claude picks and locks a personal name for every TBD boss immediately, in whatever register reads well.
• Canon support: AGAINST on every axis: [SRC: CLAUDE.md 'Cardinal rule: do NOT invent canon' + 'Naming canon (House of Velheim; patriarch Cassius Velheim)' is Josh-exclusive] [SRC: CVD §17.7 naming hard lines are Josh's] [SRC: boss_roster_rows.json binding naming-discipline clause forbids exactly this]. No canon support; listed only to reject.

Recommendation — A. Substrate-Echo methodology (the §17.7 realm-name discipline generalized)

How it fits the CVD. Option A is not a new invention — it is the discipline Josh ALREADY wrote into CVD §17.7 for the four threshold realms, generalized to boss personal names. §17.7 shows the exact mechanic four times: a documented substrate word is echoed into a coined proper noun (Ma'at -> Maatherion [SRC: CVD L1012], Tiamat -> Tiamarath [SRC: CVD L1011], Kalevala's Pohjola -> Pohljar [SRC: CVD L1010]), and tainted-uptake proper nouns (Agartha, Hyperborea, Thule, Lemuria) are dropped to reference-only [SRC: CVD L1010-1012, §19.2]. §17.1 supplies the two guardrails the method needs: 'own self-naming' (names come from the culture's own onomastics, never an exoticizing register) and 'registered/protected cultural property referenced not reproduced' — so no sacred theonym, no AIATSIS/registered name, no living-person or venerated-ancestor name becomes a boss [SRC: CVD L927-929, §17.1]. It also honors §18.3's forbiddance of cultural shorthand [SRC: CVD L1148-1156]. Crucially it respects the authority hierarchy: the METHODOLOGY is a grounded reading of existing CVD discipline (safe to recommend); each actual NAME remains Josh's per the cardinal do-not-invent rule [SRC: CLAUDE.md L14-20].

How it builds canon. It converts a 75-row open naming backlog into a repeatable, auditable procedure Josh can run at cascade speed: for any boss, (1) pull the RCM per-chapter culture/language block [SRC: RCM Ch-sections, e.g. Ch 2 Flores/Nage L81-97, Ch 8 Swahili L208-231, Ch 23 Greece/Eleusis L613-637], (2) form 2-3 substrate-plausible candidates, (3) run the §17.1 exclusion filter, (4) Josh picks/rejects/overrides. It gives the future /spine-chapter skill a naming sub-step, and it pre-commits the guardrail that already saved canon once (the roster's binding naming-discipline clause). It also forward-fits the vessel structure: named vessels Ch 2-15 get culturally-grounded personal names; framework-stub operatives Ch 16-65 can stay on op-type titles or take a vessel-culture surname; VRIL-standalone manifestations (False Naga, the echoes) correctly get NO personal name because there is no person [SRC: boss_roster_rows.json philosophy 'named vessels Ch 2-15, framework-stub operatives... VRIL-standalone manifestations where canon puts no Council presence'].

Why this option. B is wrong because canon already names vessels (Sayyid Abdallah bin Rashid Ch 8, Meryre Ch 15, Arikankharor Ch 14, Raghavendra Ch 7, Oba Ayize Ch 12) [SRC: boss_roster_rows.json those rows] — a blanket no-names rule would contradict shipped canon and flatten the named-vessel/operative/manifestation gradient the roster deliberately builds. C is barred outright by the cardinal do-not-invent rule and the roster's own binding naming clause [SRC: CLAUDE.md L14-20; boss_roster_rows.json]. A is the only option that (i) matches a discipline Josh personally codified, (ii) preserves his naming authority, (iii) protects the §17.1 high-care cultures by construction, and (iv) scales across all 75 rows. The candidates below are ILLUSTRATIONS OF THE METHOD, not assignments — every one is [DECISION NEEDED - JOSH].

Three-layer fit.

• Surface (L1): L1: A personal name makes a boss a person rather than a placeholder title — the surface-track player fights 'someone', which is exactly the roster's antagonist-narrative-first aim [SRC: boss_roster_rows.json philosophy]. The name must never signal the substrate thesis; it is just a plausible name in-world.
• Mid (L2): L2: The depth-track player who decodes inscriptions and reads the journal can recognize that a vessel's name sits correctly inside its culture's real onomastics (a Swahili/Arabic name on the Ch 8 coast, an Amarna-Egyptian name at Ch 15) — the name rewards the same documented-worldview investigation the layer is built on, without ever appropriating a real sacred name [SRC: CVD L1140 Layer 2; §17.1 L927-929].
• Deep (L3): L3: Names must NOT surface the convergence thesis or hint the Grand Sage — no name is a wink. The Architect's Ch 69 vessel-name (Maatherion is the REALM, per §17.7 L1012; the Architect himself is separate) especially must not pre-declare identity, and no boss name may hint Grand Sage identity before Ch 76 [SRC: CVD §18.1 L1132, §18.3 L1152]. The methodology's silence discipline is itself the L3 safeguard.

Test (consistency). Josh-side review, then a per-name exclusion audit before any name is locked. Verify each proposed name against: (1) it is NOT a real deity/theonym, venerated ancestor, living or historical person, or registered/protected cultural property [SRC: CVD §17.1 L927-929]; (2) it is NOT a §19.2 reference-only tainted proper noun [SRC: CVD L1010-1012]; (3) it sits inside the chapter's documented language/onomastic tradition per the RCM culture block [SRC: RCM per-chapter sections]; (4) it does not hint the thesis or Grand Sage identity [SRC: CVD §18.3 L1152]; (5) it does not collide with a locked §17.7 name or an existing canon character. A name passes only if all five hold AND Josh approves it. _Expected:_ Every locked boss name: substrate-plausible, non-appropriative, thesis-silent, Josh-approved. No sacred/registered/real-person name enters the roster.

Adversarial review — HOLDS. I independently read the cited canon rather than trusting the anchors, and the recommendation survives adversarial review. VERIFIED: (1) CVD §17.7 L1009-1012 — the echo-not-appropriate mechanic is real and shown four times (Pohjola->Pohljar, Tiamat->Tiamarath, Ma'at->Maatherion) with tainted proper nouns (Agartha/Thule/Hyperborea/Lemuria) dropped to reference-only per §19.2. (2) CVD §17.1 L927-929 — "Each culture appears with its own self-naming" and "Registered and protected cultural property is referenced, not reproduced" both confirmed, plus the Ch 57 Australia no-combat/no-family-presence hard line and the Ch 29/42/57 high-care constraints. (3) CVD §18.1/§18.3 — "Pre-Ch 76 Grand Sage identity hints are forbidden" confirmed verbatim, grounding the L3 safeguard. (4) The roster's binding naming clause is verbatim as cited: "every TBD-stub boss carries only its canon descriptive title; personal names remain [DECISION NEEDED - JOSH] per the naming-canon rule" — this bars Option C and mandates the conservative disposition. (5) The seven vessel names the "AGAINST B" argument depends on (Datu Siahaan Ch 5, Raghavendra Ch 7, Sayyid Abdallah bin Rashid Ch 8, Oba Ayize Ch 12, Arikankharor Ch 14, Meryre Ch 15, Li Si Ch 21) are NOT roster inventions — they appear in ACTIVE T4 (§7 per-chapter catalog, lines 247-305) and RCM, tracing to T3_Supporting/T3_Core. So "canon already names vessels" is factually true and Option B genuinely contradicts shipped canon. (6) T4 confirms the Architect is a role across generations (different body Ch 17 vs Ch 69), owned by T3_Core §4, and Inner Council names tie to §17.7 numerology (master number 11) — so the recommendation correctly routes Ch 55/65/69 names to Josh as deep identity canon. WHAT BROKE (nothing material): the high-care chapters Ch 10/29/42/57 are all environmental/relief-environmental with no personal-named boss, so the methodology protects §17.1 by construction — the recommendation does not smooth over any hard line, it structurally honors them. The disposition (await_josh_confirm, zero names applied, only a methodology note + skill sub-step on approval) is the correct conservative ceiling for Josh-exclusive naming canon; no over-reach. CANON-FAITHFUL, anchors check out, verdict unchanged.

Stronger option the adversary surfaced: None that replaces Option A. The only improvement is a refinement WITHIN Option A: split the 75 rows into three explicit buckets in the methodology note — (a) already-canon named vessels -> AUDIT-ONLY, no derivation; (b) framework-stub operatives with a documented culture -> substrate-echo derivation candidates for Josh; (c) VRIL-standalone/environmental manifestations and §17.1 protected-culture chapters -> NO personal name by rule. This makes the procedure deterministic per-row instead of leaving Josh to re-decide the bucket each time, and prevents the audit from ever being pointed at a name canon already fixed.

Corrections to fold in:

• Scope-of-precedent nuance: CVD §17.7's echo mechanic is written for REALM names, not personal names. Option A GENERALIZES it to personal names — a defensible extension, but the framing 'the discipline Josh ALREADY wrote into §17.7, generalized' should be stated to Josh explicitly as an extension of §17.7's scope (personal names are a new application), not merely an application of an existing rule. The personal-name guardrails are actually carried by §17.1 (self-naming + referenced-not-reproduced), which the recommendation does cite; §17.7 supplies the coining TEMPLATE, §17.1 supplies the exclusion filter. Recommend Josh be told he is authorizing a scope extension so the methodology's provenance is honest.
• For the seven already-canon named vessels (Datu Siahaan/Raghavendra/Sayyid Abdallah bin Rashid/Oba Ayize/Arikankharor/Meryre/Li Si, all in ACTIVE T4/RCM), the methodology's role is CONFIRMATION/AUDIT ONLY (run the 5-point filter against the existing name), never derivation — the recommendation implies this for Ch 8 but should generalize it so Josh is not asked to 're-assign' names canon has already fixed. Note Li Si Ch 21 is a documented HISTORICAL non-Velheim figure per T4 L305, so it is a real-person name already in canon by design (historical-figure encounter), which the 5-point 'not a historical person' filter would flag — this exception (canon deliberately uses the real historical Li Si) must be recorded so the audit does not erroneously reject shipped canon.
• Add to the methodology note that environmental/VRIL-standalone bosses (Ch 10/20/29/36/37/38/45/51/52/57/70 and similar) correctly receive NO personal name because there is no person — and for the §17.1 high-care/protected chapters (Ch 10 San, Ch 29 Sámi, Ch 42 Haiti, Ch 57 Australia) this is not merely optional but hard-line-required, since a named boss there would risk combat-against-community-representative framing. The recommendation states this for Ch 38 but should make the high-care linkage explicit as a binding sub-rule.

Apply actions. On Josh's confirmation of the METHODOLOGY (Option A): (1) add a 'Boss naming methodology' note to docs/proposals/ROSTER_PROPOSALS.md capturing the 4-step procedure + 5-point exclusion filter + the §17.7/§17.1 anchors, marked as the standing rule; (2) add a naming sub-step to the future /spine-chapter skill so each cascade chapter runs it; (3) leave every boss_name in boss_roster_rows.json unchanged (personal names stay [DECISION NEEDED - JOSH] until Josh rules each one). For any INDIVIDUAL NAME Josh picks from the candidates below (or supplies): update that row's boss_name only after Josh's explicit approval, run the 5-point audit, and record the approval in the commit message. NO name is applied without per-name Josh sign-off. Illustrative candidates for Josh to pick/reject (all [DECISION NEEDED - JOSH], all echo-not-appropriate) --- Ch 2 The Corrupted Hunter (Flores; Nage/Manggarai substrate, RCM L81-104): a coined Nage-adjacent personal name (e.g. 'Watu', 'Sina', 'Rangga' as illustrative shapes) — NOT a Nage clan-ancestor name; exclusion-check against Forth's Nage ethnography. Ch 8 First Fully-Aware Vessel (canon already names him 'Sayyid Abdallah bin Rashid', RCM Swahili/Arabic L208-231) — ALREADY NAMED IN CANON; methodology confirms it fits (Swahili-coast Arabic onomastics); no change needed unless Josh revises. Ch 23 The Corrupted Hierophant of Eleusis (Greece; RCM L613-637): a coined Greek theophoric-free personal name in the -kles/-andros/-genes register (e.g. 'Theokritos', 'Aristodemos' as shapes) — NOT the name of any real hierophant or an epithet of Demeter/Persephone; the OFFICE (Hierophant) stays the title. Ch 38 The Azores Convergence (VRIL-STANDALONE environmental capstone, no person, T4 no-presence L339-341) — NO personal name; it is a place-echo, not a being; leave as descriptive. Ch 55 The Inner Council at the Vimana / deputies (Atacama; three sector-flavored deputies, T4 L383): Velheim operatives, so House-adjacent or vessel-culture surnames in a modern register (Josh-only, since Velheim naming is §17.7 hard-line) — recommend Josh assign these directly as they are family/Council figures, not culturally-grounded locals. Ch 65 The Gathering (Antarctica pre-realm gathering; Political-Access vessel, T4 L413): a modern-operative name, vessel-culture-dependent — flag to Josh that the vessel's cultural origin isn't fixed in the row, so the substrate is unspecified; Josh sets origin then name. Ch 69 The Final Negotiation / the Architect (Maatherion is the REALM name per §17.7 L1012; the Architect is a separate ancient entity): the Architect's personal name is deep identity-canon touching the 3,000-year worldview and must NOT hint Grand Sage — recommend Josh reserve it entirely; if named, echo the Egyptian Ma'at/Duat substrate WITHOUT reusing 'Ma'at' or any real Egyptian theonym, and ensure it carries no thesis hint [SRC: CVD §18.3 L1152, §17.7 L1012].

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Every individual boss personal name is Josh-exclusive naming canon (CVD §17.7, cardinal do-not-invent rule) — the candidates above are illustrations of the methodology, NOT assignments; each requires per-name Josh sign-off before it enters boss_roster_rows.json.
• [DECISION NEEDED - JOSH] Ch 55/65 vessels are Velheim/Council operatives whose names touch §17.7 family naming and whose vessel-culture origin is not fixed in the row — Josh must set the cultural origin before a substrate-grounded name is even derivable.
• [DECISION NEEDED - JOSH] The Architect's personal name (Ch 69) is deep identity-canon with a Grand-Sage-hint hazard — recommend Josh reserve it entirely rather than derive a candidate.

---

18. Ch 58 non-anchor boss; Tier-A anchor #2 (Ch 23); Ch 43 region label

• Key: D-9 Type: assignment Disposition: await_josh_confirm

Options

• D-9 / Opt-A — Confirm Ch 58 non-anchor act-climax boss (Antarctic Rift Facility) — Ch 58 is a full chapter peak with a boss, but NOT one of the five Tier-A spine anchors. It hosts the roster's Antarctic Rift Facility: a four-phase three-tier facility part-break that recombines the rift-chain teaching nodes (46 astral / 54 layers / 56 coupling), a largest-modern manifestation, facility purge, and secondary-quantum-rift closure. Cassius absent (reserved Ch 74).
• Canon support: FOR: [SRC: T4 §2.4 L75] 'The family's operational forces are directly engaged for the first time at Ch 58 Antarctica' — a canonical peak event. [SRC: CVD §16.1 L865] every chapter has a peak, 'a boss encounter, a transformation moment...'. [SRC: T4 §8 L389] three-tier facility + largest modern installation + Cassius absent. [SRC: T4 §9.3 L393-395] rift-chain endpoint 46->54->56->58; Phase-3 stabilization closes the rift. [SRC: VRIL L332-334/L360] largest manifestation + facility purge (lineage step 2). [SRC: T1_Combat L207] Entu three-phase blended template (Pillar 8). NON-anchor status: [SRC: T1_Combat L211-215] anchors = Ch 8/23/38/55/65, Ch 58 absent; [SRC: T4 L601] same five; former 'fifth anchor' sentence was a confirmed error, deleted commit 7e23381 (GF-6/GF-7, CORPUS_SWEEP §3).
• D-9 / Opt-B — Ch 58 carries no dedicated boss (peak resolves as pure environmental/traversal set-piece) — Treat Ch 58 as a facility-infiltration + rift-closure environmental peak with no named boss entity, on the reading that the 'family does not understand the Vimana' and Cassius is absent, so there is no antagonist to fight.
• Canon support: AGAINST (weak): [SRC: T4 §2.4 L75] explicitly frames Ch 58 as the FIRST DIRECT operational-force engagement — the opposite of no-combat. [SRC: CVD §16.1 L869] 'The chapter never coasts on one mode (Pillar 11)'; [SRC: CVD §16.2 L877] combat absence 'creatively fatal.' A modern-world direct-engagement climax with no boss under-delivers the arc's largest installation. The environmental layer is a PHASE of the boss (rift closure), not a substitute for it.
• Anchor#2 / Opt-A — Build anchor #2 at Ch 23 (Corrupted Hierophant of Eleusis); 5I lock noted as residual gate — Author the mid-Tier-2 anchor at Ch 23, matching the roster's Corrupted Hierophant capstone, while flagging that the formal 'Ch 23 or Ch 25' lock nominally waits on the Session 5I T1_Trades_Master trade-arc populate.
• Canon support: FOR: [SRC: T4 L601] anchor set stated as 'Ch 8 plus Ch 23 plus Ch 38 plus Ch 55 plus Ch 65' — T4 already resolves the ambiguity toward Ch 23. [SRC: boss_roster_rows.json Ch 23 row] builds Ch 23 with three win-routes + canonical school-questline bypass [SRC: T4 L309]. RESIDUAL: [SRC: T1_Combat L212] 'Ch 23 or Ch 25 ... locks at Session 5I T1_Trades_Master populate' — the formal gate is not yet closed; Josh's ruling closes it or defers it.
• Anchor#2 / Opt-B — Defer to the Session 5I trade-arc lock; leave anchor #2 unbuilt — Hold anchor #2 unassigned until the T1_Trades_Master 5I populate mechanically resolves whether the trade questline pacing gate lands at Ch 23 or Ch 25.
• Canon support: FOR: [SRC: T1_Combat L212] the lock is explicitly deferred to 5I. AGAINST: [SRC: T4 L601] already lists Ch 23 as canonical; deferring blocks the spine cascade on a gate T4 has effectively pre-decided. Ruling 4 (stay autonomous / resolve don't defer) favors ratifying Ch 23 now with the 5I gate as a formality.
• Ch43 / Opt-A — Umbrella label 'Southeast USA (Appalachia + Sea Islands Gullah Geechee)' — Rule the canonical region a composite: Southeast USA umbrella spanning BOTH the Appalachian opioid/folk-medicine culture and the Sea Islands Gullah Geechee culture. Reconciles T4, WGM, and RCM without inventing geography.
• Canon support: FOR: [SRC: RCM L4552] section title 'Ch 43 Southeast USA — Appalachian and Sea Islands Gullah'; [SRC: RCM L1117-1122] Tier-1 substrate covers BOTH opioid/Appalachia (Quinones, Macy Dopesick, Keefe, Cavender Folk Medicine in Southern Appalachia) AND Gullah (Turner Africanisms in the Gullah Dialect); [SRC: RCM L4590-4596] search terms span opioid+Appalachia AND Gullah Geechee+sweetgrass. [SRC: T4 L357] 'Southeast USA (Appalachia) ... PHARMACEUTICAL ... opioid crisis' = the Appalachian half. [SRC: WGM L578-585] 'Southeast USA / Gullah Geechee ... Sea Islands ... opioid supply chain' = the Gullah half + shares the opioid op-type. Neither doc is wrong; each names one sub-region of the composite.
• Ch43 / Opt-B — Pick one sub-region as canonical (either Appalachia OR Gullah Geechee) — Resolve the conflict by selecting a single dominant sub-region and demoting the other to cross-reference — e.g., canonize WGM's Gullah Geechee/Sea Islands and treat Appalachia as adjacent substrate, or vice versa.
• Canon support: AGAINST: [SRC: RCM L4552 + L1117-1122] the authored substrate is deliberately BOTH cultures at Tier-1 mandatory depth; dropping either discards canonical research corpus. [SRC: WGM L585] the opioid op-type (T4's Appalachian anchor) sits inside WGM's Gullah entry — the two are already fused in canon. Choosing one manufactures a conflict the RCM does not have. Chapter-assignment/region is do-not-invent canon, so any narrowing is Josh's alone; but the evidence points to composite, not either/or.

Recommendation — CONFIRM all three: (1) Ch 58 hosts a NON-anchor act-climax boss = the Antarctic Rift Facility; (2) build anchor #2 at Ch 23 (5I lock noted as residual gate); (3) Ch 43 canonical label = "Southeast USA (Appalachia + Sea Islands Gullah Geechee)".

How it fits the CVD. All three fits serve the CVD spine and pillars without inventing canon. Ch 58 as a blended-mode boss satisfies Pillar 8 (every boss blends combat+weapons+abilities+puzzles+character; [SRC: T1_Combat L207 Entu template]) and CVD §16.1's every-chapter-peak rule [SRC: CVD L865], while honoring the §17.6 escalation principle that stages operational-tier engagement at Ch 58 before leadership-tier at Ch 74 [SRC: T4 §2.4 L75-77]. Keeping Ch 58 a NON-anchor preserves the numerical/architectural integrity of the five-anchor spine [SRC: T1_Combat L211-215; T4 L601] — the correction that already removed the false fifth anchor (GF-6/GF-7). Anchor #2 at Ch 23 aligns COMBAT and T4 [SRC: T4 L601] and lets the spine cascade proceed. The Ch 43 composite label honors CVD §17.1 cultural-care discipline by preserving BOTH documented cultures' primary-source substrate at authored depth rather than erasing one.

How it builds canon. Confirming these three stands up the Ch 43 and Ch 58 boss_roster rows for T0_Boss_Encounter population (flagship P4/P8 registry currently 0 rows per Known-issue #4), closes two of the roster's carried flags, and resolves two CORPUS_SWEEP §4 decisions (D-9 = item 9; the Ch 43 label = the roster's in-row [DECISION NEEDED]). Ch 23 ratification unblocks the mid-Tier-2 spine anchor for the Phase 6 cascade. The Ch 43 umbrella label removes a live T4-vs-WGM contradiction and pins a region descriptor the spine, WGM reconciliation, and RCM all consume. Boss personal NAMES remain [DECISION NEEDED - JOSH] (naming canon) — unaffected here; only the descriptive-title rows advance.

Why this option. Opt-A wins each sub-decision because it is the only reading the primary domain doc directly states. For Ch 58, T4 §2.4 L75 canonizes 'first direct engagement' — Opt-B (no boss) contradicts the domain owner and CVD §16.2's combat-absence-is-fatal rule. For anchor #2, T4 L601 already lists Ch 23; Opt-B (defer) blocks the cascade on a gate T4 has effectively pre-decided, against Ruling 4 (resolve don't defer). For Ch 43, the RCM authored BOTH cultures at Tier-1 depth (L4552, L1117-1122), so Opt-B (pick one) discards canonical substrate and manufactures a conflict the RCM does not have; Opt-A reconciles all three docs and invents nothing.

Three-layer fit.

• Surface (L1): L1 — Ch 58 delivers a playable, action-packed climax boss (four-phase facility part-break -> manifestation -> purge -> rift closure), satisfying the hard-line 'action-packed everywhere' ruling and Pillar 8 blended-mode. Ch 23 delivers the mid-game capstone exam. Ch 43's surface is a modern investigative infiltration boss (opioid distribution chain) grounded in a real dual-culture Southeast-USA setting.
• Mid (L2): L2 — At Ch 58 the investigation layer (rift-chain apparatus 46->54->56->58, primary-vs-regional installation architecture) becomes readable through journal/Grimoire cross-links; the family's mis-understanding of the Vimana is investigatable. Ch 43's L2 surfaces the opioid supply-chain documentation (Dopesick/Empire of Pain) and the Gullah linguistic-retention + Appalachian folk-medicine substrate as parallel documented worldviews.
• Deep (L3): L3 — Ch 58 is a Pattern node: the family's institutional apparatus is downstream of a consciousness predating it [SRC: T4 §2.5 L81]; the rift-closure is the metaphysical-as-real operating mechanically, never declared. Ch 43's L3 is convergence — West African traditions surviving 300+ years in Gullah retention alongside Appalachian three-culture syncretism, the thesis-as-convergence surfaced through Pattern, never spoken.

Test (consistency). Before populating rows: (1) confirm anchor set stays Ch 8/23/38/55/65 across T1_Combat L211-215 and T4 L601 (Ch 58 NOT added); (2) confirm the false fifth-anchor sentences remain deleted (GF-6 T3_Supporting L703; GF-7 T4 L389 — verify not reintroduced); (3) confirm Ch 43 label edit is reflected in T4 L357, WGM L578 heading, and the boss_roster Ch 43 row, and cross-checks RCM L4552; (4) run harness run_gates to confirm no pacing/FK regression when Boss_Encounter rows populate. _Expected:_ Five-anchor set intact; no reintroduced fifth-anchor sentence; Ch 43 label consistent across T4/WGM/RCM/roster; gates green (no new R1/R2/R3/FK failures).

Adversarial review — HOLDS. I independently read every cited anchor and tried to break all three sub-decisions. The core survives.

CH 58 NON-ANCHOR BOSS (Opt-A): Confirmed. T4 §2.4 L75 says verbatim "The family's operational forces are directly engaged for the first time at Ch 58 Antarctica," staging operational-tier at 58 before leadership-tier (Cassius) at Ch 74. VRIL L332 confirms Ch 58 = "negative-consciousness largest-modern manifestation in the mortal-world arc, exceeding Ch 06 False Naga and Ch 17 first Architect contact" and L360 confirms "Ch 58 Antarctica facility-purge application" — STRONGER support than the recommendation cited. RCM L5330 additionally gives Ch 58 "Inner Council escalation canonical primary" + "Third Transformation overview-effect peak." Opt-B (no boss) contradicts the domain owner and CVD §16.2 L877 "Combat absence is creatively fatal." Ch 58 is NOT a Hard Line/high-care chapter — CVD §16.2 L875 names the trio as Ch 38/57/69 — so no cultural-protection barrier. Non-anchor status is airtight: COMBAT L211-215 and T4 L601 BOTH list exactly Ch 8/23/38/55/65 (Ch 58 absent); false fifth-anchor sentences deleted (commit 7e23381, GF-6/7) and I verified they remain deleted in T3_Supporting L703 and T4.

ANCHOR #2 AT CH 23 (Opt-A): Confirmed. T4 L601 lists Ch 23; COMBAT L212 leaves the formal gate 'Ch 23 or Ch 25 ... locks at Session 5I' but T4 pre-resolves to Ch 23, and T4 L309 supplies the Ch 23 Eleusinian initiatory-corruption boss with questline bypass. Ratifying now with 5I as a formality fits resolve-don't-defer. Correctly does NOT auto-edit COMBAT L212 (a Josh-authored change).

CH 43 COMPOSITE LABEL (Opt-A): Right reconciliation, slightly misframed evidence. RCM L4552-4556 authors an explicit DUAL-ARC: 'Appalachian corridor PRIMARY plus Sea Islands Gullah Geechee SECONDARY,' both at Tier-1 mandatory depth. So Opt-B (pick one) discards authored canon. Umbrella label warranted; disposition correct because chapter region-label is do-not-invent canon.

DEFECTS (none fatal): (1) The cvd_fit cites 'CVD §17.6 escalation principle' — I read CVD §17.6 (L983): it is 'Trade and Side Questline hard lines,' NO escalation principle. The phantom citation originates in T4 §2.4 L77 and the recommendation echoes it; content is canonical, label is a live mis-cite. (2) WGM's Ch 43 geography fields (location/coords/biome, L578-585) are PURELY Gullah/Sea-Islands — only the op-type is shared with T4's Appalachia, so 'already fused in canon' overstates. (3) RCM ranks Appalachia PRIMARY / Gullah SECONDARY (L4556); the recommendation frames them as co-equal — the proposed ordering happens to match primacy, so apply-action stands.

Canon-faithfulness flags raised:

• INHERITED CITATION ERROR (not fabricated, but repeated uncritically): cvd_fit cites 'CVD §17.6 escalation principle [SRC: T4 §2.4 L75-77]'. CVD §17.6 (L983) is titled 'Trade and Side Questline hard lines' with no escalation principle. The phantom §17.6-escalation citation originates in T4 §2.4 L77; the recommendation passes the T4 error through as grounding. Escalation CONTENT is genuinely canonical (T4 §2.4 + VRIL L360 purge lineage + RCM L5330), so no world-canon is invented — but the §17.6 label should be flagged for correction, not used as support.
• OVERSTATED EVIDENCE (Ch 43): recommendation claims 'the opioid op-type sits inside WGM's Gullah entry — the two are already fused in canon.' WGM L578-585 Ch 43 geography (location=Sea Islands SC/GA; coords 32.0N 80.5W; biome coastal-lowlands/marsh) is PURELY Gullah; WGM carries the opioid op-type but zero Appalachian geography. Only op-type is shared, not region.
• UNDERSTATED PRIMACY (Ch 43): recommendation frames Appalachia/Gullah as co-equal composite. RCM L4556 ranks them Appalachia PRIMARY / Gullah SECONDARY. Ordering in the proposed label matches primacy, so harmless, but the reasoning should credit RCM's existing hierarchy rather than presenting a flat both/and.

Stronger option the adversary surfaced: none — Opt-A is correct on all three sub-decisions. The only improvement is grounding, not option choice: the Ch 58 support is under-cited and should add VRIL L332 (largest-modern manifestation exceeding False Naga and Ch 17) and RCM L5330 (Inner Council escalation canonical primary; Third Transformation peak), which independently and more decisively defeat Opt-B than the T4-only anchors the recommendation leans on.

Corrections to fold in:

• Flag that 'CVD §17.6 escalation principle' is a mis-citation inherited from T4 §2.4 L77 (CVD §17.6 = 'Trade and Side Questline hard lines'). Open a grounded-fix ticket to correct T4 L77's citation; do not rely on §17.6 as grounding for D-9.
• State that RCM L4556 ranks Appalachia PRIMARY / Gullah SECONDARY (not co-equal), which independently supports the proposed 'Appalachia + Sea Islands Gullah Geechee' ordering and strengthens the umbrella label over a flat-composite framing.
• Note WGM's Ch 43 body fields are Gullah/Sea-Islands-only; editing only the L578 heading leaves location/coords/biome Gullah-only. Flag for Josh whether the WGM body should gain an Appalachian coordinate/biome line (37.8N 82.3W Pikeville per RCM L4556) or the heading-only edit is an intentional pointer to RCM's dual-arc.
• Strengthen the Ch 58 grounding by adding VRIL L332 (largest-modern manifestation exceeding False Naga and Ch 17 contact) and RCM L5330 (Inner Council escalation canonical primary; Third Transformation peak) — both defeat Opt-B more decisively than the T4-only citations.

Apply actions. On Josh's confirm: (1) D-9 — no canon-text edit needed to CONFIRM Ch 58 as a non-anchor boss (the false fifth-anchor sentences are already deleted, commit 7e23381); mark the boss_roster_rows.json Ch 58 row's D-9 contingency as RESOLVED-confirmed, ready for T0_Boss_Encounter population as kind='chapter' NON-anchor. (2) Anchor #2 — ratify Ch 23; update the boss_roster Ch 23 row FLAG note from 'carried' to 'ruled Ch 23 (T4 L601); 5I lock a formality'; no COMBAT-doc edit unless Josh wants L212 rewritten to 'Ch 23' (that would be a Josh-authored canon edit). (3) Ch 43 label — apply the umbrella label 'Southeast USA (Appalachia + Sea Islands Gullah Geechee)' via critic-gated edits: T4 L357 'Southeast USA (Appalachia)' -> 'Southeast USA (Appalachia + Sea Islands Gullah Geechee)'; WGM L578 heading 'Ch 43 — Southeast USA / Gullah Geechee' -> 'Ch 43 — Southeast USA (Appalachia + Sea Islands Gullah Geechee)'; update boss_roster Ch 43 row to drop the [DECISION NEEDED] and cite the ruled label. All edits are region-label/data-model instantiation; run fresh-context critic + run_gates before commit. Boss NAMES stay [DECISION NEEDED - JOSH].

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Boss personal NAMES for the Ch 43 Distribution Chain vessel and the Ch 58 facility manifestation remain Josh-exclusive naming canon — rows carry descriptive titles only until Josh assigns names.
• [DECISION NEEDED - JOSH] Anchor #2: if Josh wants T1_Combat L212 physically rewritten from 'Ch 23 or Ch 25' to 'Ch 23', that is a Josh-authored canon-text change (this recommendation ratifies the assignment but does not auto-edit the COMBAT spec).
• [DECISION NEEDED - JOSH] Ch 43 exact umbrella wording is Josh's to finalize — 'Appalachia + Sea Islands Gullah Geechee' is the evidence-backed proposal; the precise phrasing/ordering is a naming-adjacent call.

---

19. The 12 open corpus-sweep decisions (CORPUS_SWEEP.md §4) — per-item reasoned recommendations

• Key: CORPUS_SWEEP_S4_12_DECISIONS Type: assignment Disposition: await_josh_confirm

Options

• Per-item slate (recommended): resolve all 12 with the per-item recommendations below — D-1 remap the six Ch-39 familiar anchors to Ch 59 Africa / Ch 60 Asia / Ch 63 Pacific revisits (Ora's bond-formation → Ch 63 Pacific, its natural Flores revisit home), specific chapter pinned by Josh/Phase-6; D-2 relocate Trade-1 Grand Communicator + Ch-39 catalog entry to the Ch 59 Africa revisit (per T1_Trades_Master §3.1 Phase-6 pin); D-3 STRIP the false CVD §3-Component-7 attribution from the 222 count and re-home it under Pillar 12 / the doc's own Beastiary framework, and let Josh rule the 222-vs-400+ reconciliation; D-4 move Ana recurrence-2 to an Antarctica chapter (Ch 58 or Ch 65), Josh picks; D-5 fix the DESCRIPTOR not the number at T4 §9.3 L393 (Ch 46 = Southwest USA rift install, delete 'Mexico (Yucatan)'); D-6 qualify §6.3 L201 to 'only Cassius/Sector-1 direct engagement' and fix the phantom 'Vril sector head' at §9.2 L383; D-7 Josh sweeps the ~8 CVD-body 'Ch 69 Agartha' residues → Maatherion (CVD-only edit); D-8 reword CVD §13.4 to a forward-reference or accept as standing pointer (CVD-only); D-9 folds into the ruled boss-roster work order (Ch 58 non-anchor boss allowed; Ch 65 = the fifth anchor); D-10 Josh confirms/cites the '18-chapter hard-line spine' count; D-11 record a supersession disposition for T99_CI_Registry in the operating-model docs (no _source edit); D-12 fold 66-vs-67-vs-69 into the standing Region_Index numerical ruling.
• Canon support: [SRC: WGM v2.3 L539 Ch 39=West Africa Modern; L747-803 Ch 59 Africa / Ch 60 Asia / Ch 63 Pacific revisits] [SRC: T1_Trades_Master §3.1 L55 Trade-1 modern home Ch 59-65 Africa, Phase-6 pin; §3.11 L93-95] [SRC: CVD §3 Component 7 L93 'Metaphysical real: the substrate operates' — no Beastiary/222] [SRC: T4 §7.1 L223 VRIL is standalone, NOT a Council sector; §9.1 L363 Ch 46=Southwest USA rift install; §6.3 L199-201 Cassius reserved Ch 74; §9.2 L383 Ch 55 = Cassius-via-deputies + two sector heads] [SRC: CVD §17.7 L1012 'Ch 69 realm name: Maatherion. Supersedes Agartha'; L1014 no-dual-references; residues L175/181/674/727/776/789/877/957] [SRC: Combat_Spec L211-215 five anchors Ch 8/23-25/38/55/65] [SRC: CVD §13.4 L722; §13.5 L727] [SRC: WGM §14.9 L1084 '66 region targets'; Region_Index CSV=67 rows; AaS L272/L460 + CLAUDE.md=69]
• Defer the ambiguous half, apply only the unambiguous grounded corrections now — Apply now only the items provable against a single unambiguous authority WITHOUT a chapter/number/name choice: D-5 (descriptor fix), D-6 (phantom Vril-sector + §6.3 qualifier), D-11 (disposition note). Route everything with a chapter-assignment, numerical, naming, or CVD-body dimension (D-1, D-2, D-3, D-4, D-7, D-8, D-9, D-10, D-12) to Josh untouched.
• Canon support: Same anchors; this option just partitions on the cardinal do-not-invent rule [SRC: CLAUDE.md L14-22] rather than proposing recommendations for the authorial half.
• Hold the entire slate for a single Josh review pass (do nothing now) — Treat all 12 as one batch for Josh, apply nothing in-session. Safest against drift but leaves grounded, single-authority corrections (D-5, D-6) unfixed and lets the false CVD attribution (D-3) and phantom sector (D-6) keep misrouting readers.
• Canon support: Conservative reading of CLAUDE.md L59 'apply on Josh's confirm, in-repo only'; but under-applies vs the already-delegated grounded-correction precedent [SRC: build/ops/preamble_trim/originals/CLAUDE_project_ORIGINAL.md L137 '~65 grounded error-corrections applied in-place (all critic-gated)'].

Recommendation — Per-item slate (recommended): resolve all 12 with the per-item recommendations below

How it fits the CVD. Every recommendation defends a CVD lock rather than inventing one. D-1/D-2 restore the geography lock (WGM §6 L539: Ch 39 = West Africa Modern; the six Indonesia/South-Asia familiar anchors and Trade-1 home belong at the Ch 59-63 revisits per T1_Trades_Master §3.1) — this protects Pillar 5 cross-cultural-link integrity, since a familiar mis-anchored to the wrong region corrupts its thread/link graph. D-3 removes a FALSE CVD attribution: CVD §3 Component 7 (L93) is 'Metaphysical real: the substrate operates' and contains no Beastiary or 222/200+ figure — leaving the mis-citation live would let a downstream consumer treat a doc-derived number as cardinal CVD canon, exactly the drift the cardinal rule guards against. D-6/D-7 defend two explicit CVD locks against the docs' own residue: §7.1 L223 states VRIL is a standalone type that does NOT map to a Council sector (so §9.2's 'Vril sector head' is a phantom), and §17.7 L1012-1014 hard-locks Maatherion over Agartha with an explicit no-dual-references rule that the CVD body itself violates ~8 times. D-9 executes Josh's standing Ruling-4 boss-roster work order. D-10/D-12 protect §17.12 numerical canon by refusing to silently ratify an uncited '18' or reconcile 66/67/69 without Josh.

How it builds canon. Clearing these 12 hardens the foundation the Phase-6 spine cascade consumes: the familiar/Trade-1 anchors (D-1/D-2) get pinned into the Ch 59-63 revisit chapters the cascade will author anyway (T1_Trades_Master §3.11 already declares those as Phase-6 pinning gates), so this decision instantiates a data-model slot rather than inventing content. D-3 gives the 222-vs-400+ figures a single honest home (catalog-entries vs tameable-species-roster are genuinely two counts). D-6 collapses an internal T4 contradiction that would otherwise propagate into every modern-world boss/operational-type row. D-7/D-8 close the last Agartha-naming and Tier-B-DLC-pointer residues so the CVD body stops contradicting its own §17.7 lock and §13.4 pointer. D-9 feeds directly into the already-approved 75-boss roster proposal (Ch 58 non-anchor boss + Ch 65 fifth-anchor). D-12 is the last facet of the standing Region_Index open question — resolving it unblocks the 69-region page build.

Why this option. The per-item slate beats 'defer the ambiguous half' because deferring gives Josh no recommendation on the nine authorial items — and Josh's working style (MEMORY: 'resolve don't defer', 'adversarial full-canon review') wants a best-supported recommendation on each, flagged as his call, not a bare punt. It beats 'hold everything' because D-5 (Ch 46 descriptor), D-6 (phantom Vril sector + §6.3 qualifier), and D-11 (disposition note) are provable against a SINGLE unambiguous authority with no chapter/number/name choice — the exact grounded-correction class already applied ~65 times this cycle under critic gate — so holding them just lets known errors keep misrouting readers. The slate's discipline is the split itself: apply_now only the three that invent nothing; every item touching a chapter assignment (D-1/D-2/D-4), a number (D-3/D-10/D-12), a name (D-7), or the CVD body (D-7/D-8) is marked await_josh_confirm with a best-supported recommendation, never silently resolved.

Three-layer fit.

• Surface (L1): L1 is untouched by all 12 as data-model/citation hygiene — EXCEPT D-1/D-2/D-9, which are player-visible: a familiar that reappears at the wrong region (Khan at 'Ch 39' when Ch 39 is Nigeria, not Sumatra) would break the surface promise that the modern-arc revisit recognizes the childhood bond; fixing the anchor to the true revisit chapter (Ch 60 Asia for Khan/Sumatra) keeps the recognition beat coherent. D-9's Ch 58 boss keeps L1 action-packed per Josh's Ruling-1.
• Mid (L2): L2 is the layer most protected here: the Grimoire/journal cross-cultural link graph, the Trade-home-base investigative trail, and the Beastiary catalog are all L2-investigable surfaces — a mis-homed Trade-1 base (D-2), a phantom 'Vril sector' (D-6), or a false CVD-Component-7 citation (D-3) would each hand the depth player a documented worldview that contradicts itself. These fixes keep L2 internally consistent.
• Deep (L3): L3 is served by D-7: the Agartha→Maatherion reframe (CVD §19.6) exists precisely to keep the Ch-69 realm's substrate Frame-A-clean (Egyptian Ma'at primary corpus) rather than carrying Theosophical occult-society load; residual 'Agartha' uses in the CVD body leak the deprecated proper-noun that the reframe was designed to retire, subtly muddying the convergent-tradition Pattern the deep player reads. Clearing them keeps the never-spoken thesis substrate clean.

Test (consistency). For the three apply_now items, verify against the cited single authority before critic-gating: D-5 — T4 §9.1 L363 reads 'Ch 46 Southwest USA... classified rift technology installation' (so §9.3 L393 'Ch 46 Mexico (Yucatan)' descriptor is the error, chapter number 46 is correct per the L528 chain 'Ch 46 → 54 → 56 → 58'); D-6 — T4 §7.1 L223 reads 'VRIL standalone operational type does not map to a Council sector' (phantom 'Vril sector head' at §9.2 L383) and §6.3 L201 must be qualified to Cassius/Sector-1 since §9.2 L383 places two OTHER sector heads on-site at Ch 55; D-11 — confirm no _source edit, only a disposition line in docs/OPERATING_MODEL_REVIEW.md. Then run harness/run_gates.py green with docs/fidelity_baseline.json refreshed if any registry cell moves. For the nine await-Josh items, the test is Josh's ruling; on D-1/D-2 re-verify the chosen revisit chapter against WGM §7 L747-803 region composites (Africa 59 / Asia 60 / Pacific 63) before writing. _Expected:_ apply_now trio verified against single authorities and critic-gated; gates green. The nine authorial items carry a recommendation + [SRC] anchors and a Josh ruling before any edit.

Adversarial review — HOLDS. Refutation attempt: I independently re-read every load-bearing anchor rather than trusting the recommendation's citations. All checked out. CVD §3 Component 7 (L93-95) is verbatim 'Metaphysical real: the substrate operates' with no Beastiary/222 count — so T3_Creatures L9's 'per CVD §3 Component 7 ... 222 total entries' is a genuine false attribution (D-3 grounded). CVD §17.7 L1012 locks Maatherion over Agartha and L1014 bars dual references; I grepped the CVD body and confirmed all 8 'Ch 69 Agartha' residues at L175/181/674/727/776/789/877/957 (D-7 real, CVD-body-only, Josh-exclusive — correctly routed). CVD §17.1 L933 is the scoped 'no combat against community representatives' (GF-1 correct). Combat_Spec L211-215 lists exactly five anchors Ch 8/23-25/38/55/65 with Ch 58 absent (D-9/GF-6/GF-7 grounded). T4 §7.1 L223 states VRIL 'does not map to a Council sector,' making the 'Vril sector head' at §9.2 L383 a genuine phantom (D-6 grounded). T4 §9.3 L393 verbatim 'Ch 46 Mexico initial apparatus deployment (Yucatan installation)' contradicts §9.1 L363 'Ch 46 Southwest USA' and §9.2 places Ch 47=Mexico, Ch 48=Central America — D-5 descriptor fix grounded, fix direction (keep ch 46, fix descriptor) correct. T4 §6.3 L201 'only direct Council-member engagement' is genuinely contradicted by L383 (sector heads in on-site direct contest at Ch 55) — D-6 §6.3 qualifier grounded. Trades §3.1 L55 pins Trade-1 modern home to 'Ch 59-65 continental revisit Africa, specific chapter at Phase 6 spine cascade pinning' (D-2 authorial, correctly routed). WGM L539 = 'Ch 39 — West Africa Modern' (D-1 remap warranted). WGM L1084 = 'Sixty-six region targets' (D-12 66/67/69 spread real). 'Eighteen'/'18-chapter' has zero matches in CVD (D-10 uncited, correctly routed). D-11 target docs/OPERATING_MODEL_REVIEW.md exists and the disposition is note-only, no _source edit. Attack results: (1) invented-canon-as-grounded — NONE; every apply_now item is provable against a single authority, and every chapter/number/name pick is explicitly flagged do-not-invent with Josh ruling. (2) authority-hierarchy — clean; every T-doc claim was checked against CVD, and the recommendation defends CVD locks rather than letting T-docs override them. (3) hard-line/§17 smoothing — none; GF-1 restores the §17.1 scoping and the Pillar-3 hard line, and D-7 defends the §17.7 name-lock. (4) stronger option — none materially better; Options 2 and 3 under-serve Josh's resolve-don't-defer style. (5) three-layer — holds; D-1/D-2/D-9 are correctly identified as player-visible while the citation-hygiene items are honestly infrastructure. (6) too-aggressive disposition — NO apply_now item is Josh-exclusive invention: D-5 is a pure descriptor fix, D-6 splits the grounded phantom-deletion from the Josh-flagged replacement-sector choice (D-6a), D-11 is a note-only disposition. The disposition split (3 apply_now, 9 await_josh) faithfully partitions on the cardinal rule. The recommendation survives.

Corrections to fold in:

• D-1 chapter picks are shaky against WGM and Josh should be told so: WGM L757 folds 'Southeast Asia' into Ch 60 Asia revisit territory (revisit of Ch 4-7,16,20-22), while Ch 63 Pacific (L781) is composed of Zealandia + Pacific Islands + Pacific Northwest + Australia cross-ref + Oceania. The recommendation routes Kira/Bali and Ora/Flores to Ch 63 Pacific, but Bali and Flores are Indonesian/Southeast Asia, which WGM's cleaner reading assigns to Ch 60 Asia, not Ch 63. This is a weakness in the SUGGESTED target only (correctly flagged as Josh's do-not-invent call, so not a cardinal-rule violation), but the Asia-vs-Pacific split for Indonesia is itself ambiguous in WGM and Josh should be surfaced that ambiguity rather than handed a confident Ch 63 pick.
• D-6 §6.3 rewording says 'other Inner Council sector heads' (plural), but once the phantom 'Vril sector head' is deleted, L383 names only ONE valid other sector head on-site at Ch 55 (Surveillance/Military), with Banking present only via deputies. The qualifier should read singular (the Surveillance/Military sector head is the one other direct on-site engagement) pending Josh's D-6a ruling on who the third Ch-55 attendee actually is; the plural slightly overstates the verified canon.
• Minor: the recommendation's test/reasoning at one point cites the D-5 corroborating authority as '§9.1 L363' while the error itself is at §9.3 L393 — both citations are correct (L363 in the §9.1 Ch 39-50 block establishes Ch 46=Southwest USA; L393 in §9.3 carries the wrong descriptor), but the apply-text should consistently cite §9.1 L363 as the corroborating source, not a bare 'per §9.1', to avoid a reader mistaking which paragraph is authoritative.

Apply actions. SPLIT BY ITEM. apply_now trio (grounded, single-authority, critic-gated — same class as the ~65 already applied): (D-5) T4 §9.3 L393 OLD 'Ch 46 Mexico initial apparatus deployment (Yucatan installation per Ch 39-50 cascade).' → NEW 'Ch 46 Southwest USA initial apparatus deployment (classified rift technology installation per §9.1).' (D-6a) T4 §9.2 L383 OLD 'Vril sector head' → NEW 'Corporate Resource Extraction sector head' OR delete the third-member clause — FLAG to Josh which sector, since no 'Vril sector' exists (§7.1 L223); (D-6b) T4 §6.3 L201 OLD 'The Ch 74 Cassius Velheim confrontation is the only direct Council-member engagement at canonical scope.' → NEW 'The Ch 74 Cassius Velheim confrontation is the only direct engagement with the Sector-1 Central Banking head; Ch 55 Atacama is the one modern-world chapter where other Inner Council sector heads coordinate direct on-site contest per §9.2.' (D-11) add a supersession disposition paragraph to docs/OPERATING_MODEL_REVIEW.md noting T99_Custom_Instructions_Registry [ACTIVE v1.1] describes retired constructs (per-project CI, memory markers, Drive-populate completion) and must not be trusted as live operational canon — NO _source edit. Then: critic-gate all three, run harness/run_gates.py, commit with evidence. AWAIT-JOSH nine (recommendation staged, do NOT edit until ruling): D-1 remap the six familiar Ch-39 anchors (Khan→Ch 60 Asia, Bodhi→Ch 60 Asia, Desta→Ch 59 Africa, Kira/Bali→Ch 63 Pacific, Murugan→Ch 60 Asia, Ora bond-formation→Ch 63 Pacific) — Josh confirms exact chapters (chapter assignment = do-not-invent); D-2 relocate Trade-1 Grand Communicator + Ch-39 catalog entry to Ch 59 Africa revisit (Phase-6 pin per §3.1); D-3 strip 'per CVD §3 Component 7' from T3_Creatures L9/L78 (false attribution — provable) and re-home the 222 under the doc's own Beastiary framework + Pillar 12, and Josh rules 222-vs-400+ reconciliation (numerical); D-4 Josh picks Ana recurrence-2 target (Ch 58 or Ch 65 Antarctica vs keep Ch 41); D-7 Josh sweeps the ~8 CVD-body 'Ch 69 Agartha' → 'Ch 69 Maatherion' (L175/181/674/727/776/789/877/957 — CVD edits Josh-only), and Trade_Master §3.9 L87 + CVD §13.5 followers align after; D-8 Josh rewords CVD §13.4 L722 to forward-reference or accepts as standing pointer; D-9 fold into the ruled boss-roster proposal (Ch 58 non-anchor boss authored; Ch 65 written in as the fifth anchor in T4 §4.4/§9.2 + T3_Supporting §9.2 bodies); D-10 Josh confirms or cites the '18-chapter hard-line spine' count (uncited in CVD/T0_Hard_Lines); D-12 fold 66(WGM §14.9)-vs-67(Region_Index CSV)-vs-69(cardinal canon) into the standing Region_Index numerical ruling.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] D-1 (six familiar Ch-39→revisit-chapter remap) is a chapter-assignment decision — CARDINAL do-not-invent; my chapter picks (Khan/Bodhi/Murugan→Ch 60, Desta→Ch 59, Kira/Ora→Ch 63) are best-supported recommendations only, Josh rules
• [DECISION NEEDED - JOSH] D-2 Trade-1 modern-home specific chapter is an explicit Phase-6 pinning gate (T1_Trades_Master §3.11) — Josh/cascade authorial
• [DECISION NEEDED - JOSH] D-3 the 222-vs-400+ numerical reconciliation is §17.12 numerical canon — Josh-exclusive; only the false-CVD-attribution strip is grounded
• [DECISION NEEDED - JOSH] D-4 Ana recurrence-2 target chapter is a chapter assignment — Josh-exclusive
• [DECISION NEEDED - JOSH] D-7 all ~8 CVD-body Agartha→Maatherion edits are CVD-body edits — Josh-only per CLAUDE.md
• [DECISION NEEDED - JOSH] D-8 CVD §13.4 rewording is a CVD-body edit — Josh-only
• [DECISION NEEDED - JOSH] D-9 Ch 58 boss + Ch 65 fifth-anchor content is canon invention delegated by Josh's Ruling-4 boss-roster work order — proposal, not silent authoring
• [DECISION NEEDED - JOSH] D-10 the '18-chapter hard-line spine' count is uncited numerical canon — Josh confirms or supplies citation
• [DECISION NEEDED - JOSH] D-12 the 66/67/69 region-count reconciliation is §17.12 numerical canon — Josh-exclusive
• [DECISION NEEDED - JOSH] D-6a which sector head is the third Ch-55 attendee (replacing the phantom 'Vril sector head') is a canon detail — flag to Josh even though the phantom-deletion itself is grounded

---

20. SSBL Appendix B unmigrated canon — migration work-orders for Entries 18-21

• Key: ssbl_appendix_b_entries_18_21_migration Type: assignment Disposition: await_josh_confirm

Options

• A — Migrate all four per SSBL's named targets, one work-order per entry (bundled early-P6) — Four discrete critic-gated migration work-orders executed before/early in Phase 6: Entry 18 recognition-mechanism six-subtype schema → INTEG §8 worldstate structure (with a recognition_mechanism_subtype field seed in Worldstate_Variables); Entry 19 dual-anchor + forward-arc structural-antagonist-limit → T4 §8 (new §8.4 or headers on existing per-chapter lines); Entry 20 two-register trade schema → T1_Trades_Master new §5.x/§7.x; Entry 21 reputation-biography propagation existence-registration → Worldstate_Variables new WS_026 + Schema_Dictionary stub, full spec deferred post-cascade per SSBL.
• Canon support: [SRC: SSBL L1032 Entry 18 target 'T3_Core OR INTEG at recognition-mechanism subtype canonical schema']; [SRC: SSBL L1046 Entry 19 target 'T4 §8 antagonist-network operational topology']; [SRC: SSBL L1058 Entry 20 target 'T1_Trades_Master']; [SRC: SSBL L1070 Entry 21 target 'T0_Worldstate_Variables existence registration ... full spec deferred to post-cascade']. Confirmed-missing: [SRC: INTEG L31,L115 — only Ch57 elder integrity-read, no subtype schema]; [SRC: T4 §8.2 L277,L283,L285 — per-chapter limits present but no dual-anchor/forward-arc registration]; [SRC: T1_Trades_Master §7.1 L214, §8.1 L230 — substrate concepts referenced, no two-register schema]; [SRC: Worldstate_Variables CSV rows end WS_025, no reputation-biography variable].
• B — Migrate 18/19/20 as canon now; defer Entry 21 entirely to post-cascade — Same as A for Entries 18-20, but skip even the existence-registration for Entry 21 since SSBL itself defers the full spec to the post-cascade architecture-gaps queue. Entry 21 stays a queue item, surfacing only if Ch 11 build needs it.
• Canon support: [SRC: SSBL L1072 'Full system spec deferred to post-cascade architecture gaps queue ... Full spec produced post-summer book processing per Project 2 sequence canon']; [SRC: SSBL L1080 fallback — incomplete entry 'surfaces at Phase 6 Spine Cascade Stage 2 canonical question review for the chapter where the entry's canonical content first applies']. Against: SSBL L1070 still names an existence registration ('registers at canonical existence level') distinct from the deferred full spec.
• C — Do nothing pre-cascade; let all four surface via the SSBL Stage-2 fallback at Ch 11 build — Rely wholly on SSBL's own safety net: any unmigrated entry surfaces at the Ch 11 Spine cascade Stage-2 canonical-question review. No pre-cascade migration work.
• Canon support: [SRC: SSBL L1080 'If canonical migration of any entry is incomplete at Phase 6 launch, the entry surfaces at Phase 6 Spine Cascade Stage 2 ... (Ch 11 Congo for Entries 18-21 origin chapter)']. Against: CLAUDE.md Known-issue #4 wants foundation-readiness gaps closed 'before/early in the cascade'; Entry 19 forward-arc anchors touch Ch 13/22/49/57/58/69 that build across the whole arc, so deferring to a single Ch 11 review under-serves the forward regions.

Recommendation — A — Migrate all four per SSBL's named targets, one work-order per entry (bundled early-P6)

How it fits the CVD. All four are documentation/data-model instantiations of canon Josh already authored in v2.3, not new invention — they harden the substrate the Layer-2/Layer-3 experience is built on. Entry 18's recognition subtypes and Entry 19's structural-antagonist-limit are the mechanical spine under CVD §18.2 Layer-3 'convergent traditions across disconnected cultures' [SRC: task-prompt CVD §18.2 three-layers] — the Pattern is surfaced, never declared. Entry 19 directly serves CVD §17.6 escalation and the §17.1 protected-chapter hard lines (Ch 29/38/57 no-antagonist-presence) by giving those chapters a canonical structural reason the antagonist fails, consistent with [SRC: T4 §8.2 L321 Ch29 'antagonist-family operational-forms are defeasible', L387 Ch57 'no operational presence by canonical design']. Entry 20 serves CVD §14.3-14.4 trade substrate-teaching by formalizing that Mbuti community-substrate and Persia institutional-individual are structurally distinct legitimate registers [SRC: T1_Trades_Master §7.1 L214, §10.4 L279]. Entry 21 serves the CVD §13 voice / §17.9 path-variable systems by registering that reputation-biography writes are integrity/concealment-aware [SRC: Worldstate_Variables WS_025 identity_hiding_status enum already carries CONCEALMENT_WITH_HOSTILE_RECOGNITION, the exact write-target Entry 21 scopes].

How it builds canon. Stands up four load-bearing substrate pieces the cascade consumes: (1) a recognition-mechanism subtype vocabulary the per-region build selects-or-extends at Range-Review Grid 6 [SRC: SSBL L1038 'future region builds either map to existing recognition-mechanism subtype or register a new subtype']; (2) a T4 forward-arc structural-limit register that pre-declares which regions the antagonist STRUCTURALLY FAILS in, unblocking Ch 13/22/49/57/58/69 boss/no-boss decisions [SRC: SSBL L1048]; (3) a two-register trade schema that every per-trade-master authoring pass keys against [SRC: SSBL L1062 'Range Review Grid 2 Trade Questline audit criterion verifies trade-master schema applied per canonical structure per trade']; (4) a reputation-biography existence anchor (WS_026) that the Ch 11-forward Spine entries write into, with the heavy full-spec correctly deferred [SRC: SSBL L1072].

Why this option. A beats C because CLAUDE.md Known-issue #4 mandates closing foundation-readiness gaps before/early in the cascade, and Entry 19's forward anchors span the whole arc (Ch 13→69) — a single Ch 11 Stage-2 review cannot pre-inform boss decisions for Ch 49/57/58. A beats B because SSBL L1070 explicitly distinguishes an existence-registration (do now) from the full spec (defer) for Entry 21; B drops the existence anchor the Ch 11-forward Spine entries need to write against, forcing a mid-cascade stub anyway. A honors the SSBL's own target routing for every entry and keeps the expensive Entry-21 full spec deferred exactly where SSBL puts it.

Three-layer fit.

• Surface (L1): L1 SURFACE — invisible: recognition subtypes, antagonist structural-limits, trade-master schema registers, and reputation writes all operate mechanically (an elder recognizes you, a trade master teaches, the antagonist simply is-absent at Ch 57) without the player engaging the schema. Entry 19 specifically guarantees protected chapters stay playable via the environmental/structural framing rather than antagonist combat [SRC: T4 §8.2 L387 Ch57 combat 'routes through environmental hazards plus astral mode'].
• Mid (L2): L2 MID — investigatable: the recognition subtypes (empty-space, tendi-tradition, kaya-elder, land-recognition) and the structural-antagonist-limit become legible to the depth player through journal/Grimoire cross-cultural links — why the forest 'already knows' at Ch 11 versus why the land recognizes at Ch 57 [SRC: SSBL L1030 Entry 18 subtypes; L1044 Entry 19 body-and-relationship substrate]. The two-register trade schema surfaces as documented worldview behind each trade master [SRC: T1_Trades_Master §10 worldview-teaching].
• Deep (L3): L3 DEEP — the Pattern: Entry 18's convergent recognition-across-disconnected-cultures and Entry 19's 'knowledge in bodies-and-relationships is uncapturable' ARE the thesis-as-convergence — surfaced through structure, never declared. These migrations must add NO thesis statement; they register mechanism only, preserving the §17.9 show-don't-tell / no-pre-Ch-76 discipline [SRC: task-prompt CVD §18.2 L3 'the thesis is the convergence; it is never spoken'].

Test (consistency). After each work-order, verify against: Entry 18 — INTEG §8 gains a recognition_mechanism_subtype schema enumerating the six §7.5 subtypes + recognition-via-land, and Worldstate_Variables gains a subtype field, with NO thesis declaration (grep the new text for forbidden 'Grand Sage'/thesis language per §17.9). Entry 19 — T4 §8 explicitly names Ch 07 + Ch 11 as the dual-anchor pair AND registers Ch 13/22/49/57/58/69 as structural-antagonist-limit STRUCTURALLY-FAILS regions; cross-check each against its existing §8.2/§9 chapter line and the §17.1 protected-chapter hard lines (Ch 29/38/57 no-presence must stay intact). Entry 20 — T1_Trades_Master gains an institutional-individual vs community-substrate two-register schema section applied across all 10 trades, consistent with existing §7.1 Mbuti + §10.4 per-culture framing. Entry 21 — Worldstate_Variables gains WS_026 reputation_biography_propagation existence-registration whose concealment scoping matches WS_025 identity_hiding_status enum values, with full spec left in the post-cascade queue. Run the harness fidelity gate on the two edited CSVs to confirm only the intended new rows changed. _Expected:_ Four target docs each carry the entry's canonical content sourced verbatim-quotable from the frozen SSBL; harness fidelity gate shows only the intended WS_026 (+ Entry-18 subtype field) additions as new/changed cells, all other registry cells byte-exact; no §17.9/§17.1 hard-line violation introduced; a fresh-context adversarial critic passes each DRAFT→ACTIVE edit.

Adversarial review — HOLDS. I independently re-read every load-bearing anchor rather than trusting the recommendation. Verdict: it SURVIVES adversarial review.

WHAT I VERIFIED (held):

• SSBL Appendix B, L1020-1082: all four entry targets and line-anchors check out verbatim. Entry 18→INTEG-or-T3_Core (L1032), Entry 19→T4 §8 (L1046), Entry 20→T1_Trades_Master (L1058), Entry 21→Worldstate_Variables existence + post-cascade full spec deferred (L1070/L1072). The SSBL fallback at L1080 (surfaces at Ch 11 Stage-2) is real.
• Confirmed-missing claims are accurate: INTEG (L31/L115) has only the Ch57 elder integrity-read, no subtype schema. T4 §8.1-§8.3 exist but carry NO dual-anchor/structural-antagonist-limit register. T1_Trades_Master §7.1/§10.4 reference Mbuti community-substrate concepts but have no formal two-register schema. Worldstate_Variables CSV ends at WS_025; no reputation-biography variable.
• AUTHORITY HIERARCHY (the key attack): SSBL is a T99 workflow doc (CLAUDE.md: 'not world canon'). But Appendix B is explicitly a MIGRATION QUEUE for pre-authored v2.3 canonical findings (Entries 18-21, dated April 27 2026). The recommendation correctly treats SSBL as the FROZEN SOURCE to migrate content FROM into T1/T4/T0, not as a canon authority in itself. No hierarchy violation.
• CARDINAL RULE (don't invent): Entries are pre-authored; migration is structured-form work. Disposition is await_josh_confirm (not apply_now), and every discretionary choice (Entry-18 INTEG-vs-T3_Core target, Ch49-vs-50 chapter number, WS_026 field shape, Entry-20 placement) is surfaced as an open subflag rather than resolved. This is the discipline the cardinal rule demands.
• §17 HARD LINES: Entry-19's STRUCTURALLY-FAILS registration for Ch 57/38/52/29 only records the pre-existing no-presence framing (CVD L83 'every chapter Velheim presence except Ch 57'; T4 L387 'no operational presence by canonical design'). The L1-playable three-layer claim is CVD-faithful — CVD L181 explicitly rejects 'no combat' as 'creatively fatal' and L933 routes Ch 57 combat through environmental/astral mode. No hard-line smoothed over.

WHY A OVER B/C HOLDS: SSBL L1070 genuinely distinguishes an existence-registration (do now) from the deferred full spec, so B's dropping of the existence anchor forces a mid-cascade stub anyway. Entry-19 forward anchors span Ch 13→69, so C's single Ch-11 Stage-2 review cannot pre-inform Ch 49/57/58 boss decisions — CLAUDE.md Known-issue #4 mandates closing these before/early in the cascade. A is the canon-faithful choice.

Corrections below are apply-plan imprecisions the recommendation's own mandated critic gate would catch; none changes the verdict.

Stronger option the adversary surfaced: A just-in-time variant of A, sequenced by first-consuming gate rather than bundled 'early-P6': Entry 18 (recognition subtypes) and Entry 19 (structural-antagonist-limit) anchor at Ch 11/12 and Range-Review Grid 6, so they are genuinely early-cascade; but Entry 20's two-register schema is only consumed at Range-Review Grid 2 (trade-questline audit) and Entry 21 is a marker-only existence stub. Front-loading Entry 20's all-10-trades schema authoring before its consuming gate is avoidable work. Recommend Josh consider migrating 18/19 first (before Ch 11 build), Entry 21's minimal existence-stub with them (so Ch-11-forward Spine entries have a write target), and timing Entry 20 to its Grid-2 gate. This is a refinement of A, not a refutation — A as written is still defensible and honors SSBL's own routing.

Corrections to fold in:

• Entry-21 apply-plan claims WS_026 concealment scoping 'matches WS_025 enum values' including OPEN_PRESENCE. Verified: WS_025 identity_hiding_status enum is FIRST_CONCEALMENT/ACTIVE_CONCEALMENT/NEAR_REVEAL/REVEAL_RECOVERY/CONCEALMENT_WITH_RECOGNIZED_ALLY/CONCEALMENT_WITH_HOSTILE_RECOGNITION. OPEN_PRESENCE is NOT a WS_025 value — it comes from SSBL Entry-21 text (L1068), not the registry. Two of three named states match; OPEN_PRESENCE does not. The WS_026 spec must cite SSBL Entry-21 for OPEN_PRESENCE, not WS_025.
• SSBL version label mismatch: the task prompt and filename say v3.2, but the doc's own footer (L1082) reads 'End of T99_Story_Spine_Build_Logic v3.0.' Per CLAUDE.md trust-content rule this is a label lag, not a content problem, but the migration commit should note it so the frozen-source citation is unambiguous.
• Open-subflag #2 slightly overstates the Ch49-vs-Ch50 conflict. WGM L646 shows Ch 49 itself carries the 'uncontacted-peoples territorial-defense engagement at FUNAI-protocol respectful-distance hard line per CVD §17.1' (Colombia), while Ch 50 is the Amazon-basin modern home. So SSBL's 'Ch 49 Amazon uncontacted' (L1048) is most likely CORRECT for the structural-limit anchor, and the Trades 'Ch 50 Amazon modern home' is a different beat. The flag to Josh is still appropriate (numerical canon is Josh-exclusive), but should be reframed as 'confirm Ch 49 is the intended structural-limit anchor' rather than 'SSBL contradicts WGM.'

Apply actions. On Josh's confirm, execute four critic-gated work-orders (each: session edit → fresh-context adversarial critic → commit), sourcing canon verbatim-quotable from frozen SSBL Appendix B:

ENTRY 18 → INTEG (primary) + Worldstate_Variables (field seed). In INTEG add a 'Recognition-Mechanism Subtype Schema' subsection under the worldstate/recognition structure enumerating the six recognition-via-individual-perceiver subtypes (empty-space, embodied-practice, tendi-tradition, kaya-elder-protocol, liturgical-individual-perceiver, family-network) plus recognition-via-land first-instance (Ch 11 Mbuti 'the forest already knows') per SSBL L1030; note Ch 02-11 retro-applicability and the Grid-6 select-or-register-new rule per SSBL L1036/L1038. In Worldstate_Variables add a recognition_mechanism_subtype field (either a new WS variable or an extension on the region-recognition object). Rationale for INTEG-over-T3_Core: recognition is an integrity-read worldstate behavior already anchored in INTEG L31/L115, not a character trait.

ENTRY 19 → T4 §8 (new §8.4 'Structural-Antagonist-Limit Dual-Anchor Canon'). State the pattern (knowledge in bodies-and-relationships is structurally uncapturable by a text/institution-capturing apparatus) per SSBL L1044; name Ch 07 Tamil Siddha + Ch 11 Ituri as the two canonical anchors; register the forward-arc STRUCTURALLY-FAILS regions Ch 13/22/49/57/58/69 with permitted-scope per SSBL L1048; add the Ch 12 Grid-6 audit criterion per SSBL L1050. Add cross-headers on the existing §8.2 Ch 07/10/11 lines pointing to §8.4. Do NOT alter the §17.1 protected-chapter no-presence canon.

ENTRY 20 → T1_Trades_Master (new schema section, e.g. §5.x or a §7 header). Add 'Trade-Master Two-Register Schema': institutional-individual (named master, location, lineage) vs community-substrate (community-as-master, place-as-master, lineage-as-distributed-substrate) per SSBL L1056; give the canonical examples Trade 5 Alchemy Persia = institutional-individual, Trade 1 Animal Communication Mbuti net-hunting = community-substrate per SSBL L1060; add the Grid-2 audit criterion per SSBL L1062. Reconcile with existing §7.1 (Mbuti) and §10.4 (per-culture) framing.

ENTRY 21 → Worldstate_Variables (WS_026 existence-registration) + Schema_Dictionary stub. Add WS_026 reputation_biography_propagation registering canonical existence + core distinctions per SSBL L1068: per-chapter path-and-integrity-variant writes across quests/NPCs/regions, concealment-mode-aware scoping (CONCEALMENT_WITH_RECOGNIZED_ALLY→ally network, CONCEALMENT_WITH_HOSTILE_RECOGNITION→antagonist intelligence private register, OPEN_PRESENCE→public record), matching WS_025 enum values; note cross-refs to WS_025, T4 antagonist intelligence record, Entry-19 protection per SSBL L1074. Leave the FULL system spec in the post-cascade architecture-gaps queue per SSBL L1072 (do NOT author T1_Personal_Dimension_Spec now). Run harness fidelity gate on the two edited CSVs. Flag OPEN-SUBFLAG items below for Josh before committing anything that reads as new canon rather than migration.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Entry 18 target choice INTEG-vs-T3_Core: SSBL L1032 offers both; I recommend INTEG (recognition is an integrity-read worldstate, already anchored INTEG L31/L115), but Josh should confirm since T3_Core is the co-named target.
• [DECISION NEEDED - JOSH] Entry 19 forward-arc anchor Ch 49 vs Ch 50 Amazon: SSBL L1048 says 'Ch 49 Amazon uncontacted' but WGM/registries anchor Amazon at Ch 50 (Trade-4 modern home, [SRC: T1_Trades_Master §6.5 L206 'Ch 50 Amazon']). Confirm the correct chapter number before registering the structural-limit anchor — this is a numerical-canon touch reserved to Josh.
• [DECISION NEEDED - JOSH] Entry 21 WS_026 exact field shape (single existence-flag variable vs an object schema) is a data-model design choice; I propose an object mirroring WS_019/WS_020 shape, but the full spec is deferred so the minimal existence-registration form should be Josh-confirmed to avoid pre-empting the post-cascade spec.
• [DECISION NEEDED - JOSH] Whether Entry 20's two-register schema is authored as new §5.x, folded into §7, or added as a §4-adjacent schema section is a placement choice inside Trades_Master that affects section numbering — confirm placement.

---

21. Confirm the seven combat-mode slugs (dominant_combat_mode enum)

• Key: 0.D.1_combat_mode_enum Type: creative Disposition: await_josh_confirm

Options

• CONFIRM as-proposed (7 CVD §12.2 modes; slugs: direct, stealth, puzzle_routed, environmental, vril_management, familiar, astral) — Adopt the seven CVD §12.2 modes verbatim as the closed dominant_combat_mode set, with the proposed lowercase slugs. puzzle_routed stays deliberately distinct from the mode_variety_check 'puzzle' beat-axis key to prevent axis collision.
• Canon support: [SRC: CVD §12.2 L654-666] is the ONLY canon list attached to the R1/R2 rule sentence ('Two consecutive chapters never share dominant combat type. Every five-chapter window contains three or more distinct combat-mode dominant chapters,' L664). [SRC: HARDLINES row HL_0049 (CVD §17.10)] restates that rule verbatim as a hard line, using the exact phrases 'dominant combat type' / 'combat-mode dominant chapters' — binding the rule to §12.2's seven-item 'Every chapter combines:' list (L656-662): Direct, Stealth/restraint, Puzzle-routed, Environmental, Vril-management, Familiar-leveraged, Astral/dream-world. [SRC: SCHEMA (grep dominant_combat_mode/combat_mode → no matches)] confirms no pre-existing enum, so this instantiates rather than overrides canon. 7 values ≥ R2's floor of 3-distinct-per-5-window.
• RE-SLUG puzzle_routed → puzzle (collapse to the beat-axis spelling) — Adopt the same seven modes but spell the third slug 'puzzle' to match everything else's shorthand, accepting collision with the Pillar-11 mode_variety_check 'puzzle' beat key.
• Canon support: [SRC: CVD §12.2 L658] names it 'Puzzle-routed combat (boss encounters with non-combat resolution components)' — a combat mode, whereas [SRC: CVD §12.2 L652-653 / Pillar 11] 'puzzle' as a per-chapter beat is a DIFFERENT axis (mode_variety_check c/p/l/ch/t). The topography table already carries both a dominant-mode column and a p (puzzle beat) flag in 'c/p/l/ch/t' [SRC: docs/archive/TOPOGRAPHY_TABLE.md L29-109]. Collapsing the slug would make a chapter read 'mode=puzzle, p=Y' vs 'mode=puzzle, p=N', an ambiguity canon's own two-axis framing warns against. Weakly supported — canon separates the concepts.
• ADOPT T1_Combat §9.1 ten-category set instead (direct/finer-grained encounter enum) — Use the ten COMBAT §9.1 encounter categories (person-to-person, person-to-animal, group, boss, social boss, environmental, stealth, naval, mounted, vril-projection) as the dominant_combat_mode enum.
• Canon support: [SRC: COMBAT §9.1 L319-330] is finer-grained and concrete. BUT [SRC: COMBAT authority note L472-489] subordinates T1_Combat to CVD, and the R1/R2 rule lives in CVD §12.2 / HL_0049 over the SEVEN modes, not these ten. These ten are 'encounter categories' (a texture taxonomy), not the 'dominant combat type' the rule counts. Several (naval, mounted, group, person-to-animal) are region/setpiece descriptors that cut across the seven modes rather than being alternatives to them, and 'social boss' is explicitly 'No weapon engagement' [SRC: COMBAT L325], colliding with Josh Ruling 1 (L1 action-packed everywhere). Canon does not tie R1/R2 to this list — using it would silently redefine the hard line's counting unit.

Recommendation — CONFIRM as-proposed (7 CVD §12.2 modes; slugs: direct, stealth, puzzle_routed, environmental, vril_management, familiar, astral)

How it fits the CVD. The seven modes ARE a CVD §12.2 list; adopting them as the enum is the most literal possible reading of canon, not an invention. The R1/R2 hard line [SRC: HARDLINES HL_0049 / CVD §17.10] can only be mechanically enforced against a closed set, and CVD attaches exactly one closed set to it — these seven [SRC: CVD §12.2 L654-664]. Every mode maps to a concrete CVD-sanctioned player mechanic: environmental → the §12.3 environmental-boss protocol (L670-676) that Josh Ruling 1 designates the playable layer for protected/relief chapters; astral → §4.5 sanctuary-preservation combat; vril_management → §12.3 disciplined-vril resolution. None is combat-absent, so the set is consistent with Ruling 1 (L1 action-packed everywhere). The slug puzzle_routed preserves CVD's own two-axis separation of 'combat mode' (§12.2) from 'puzzle beat' (Pillar 11 mode-variety).

How it builds canon. Instantiates the dominant_combat_mode column enum that the 79-row topography table [SRC: docs/archive/TOPOGRAPHY_TABLE.md] and the Phase-6 spine cascade both consume, turning the R1/R2 hard line from prose into a gate-checkable constraint (already validated: R1×3, R2×1 residual in the table). It fills the SCHEMA gap (no combat_mode enum exists) with a CVD-anchored closed set, and gives HL_0049 a machine-enforceable vocabulary so future spine work cannot drift the counting unit. It also standing-resolves the axis-collision risk by locking puzzle_routed ≠ puzzle for every downstream consumer.

Why this option. Option 1 is the only choice where the enum and the hard line it enforces share the same canon list — CVD §12.2 is literally the list HL_0049's rule counts over. Option 2 (re-slug to 'puzzle') destroys CVD's own separation of the combat-mode axis from the Pillar-11 puzzle beat axis, both of which coexist in the topography row schema — a self-inflicted ambiguity. Option 3 (§9.1 ten categories) silently redefines the hard line's counting unit against a subordinate T1 doc's texture taxonomy, imports a 'social boss / no weapon engagement' category that fights Josh Ruling 1, and mixes region setpiece descriptors (naval/mounted) with true modes. CVD wins conflicts; only Option 1 keeps the enum identical to the CVD list the rule cites.

Three-layer fit.

• Surface (L1): L1 players never see the slug; they experience the consequence — no two adjacent chapters coast on the same combat texture, and every 5-chapter window mixes ≥3 textures (the enum is what makes that authorable and testable). Every mode including environmental/astral/vril_management is a playable combat layer per CVD §12.3 + Ruling 1, so 'action-packed everywhere' holds.
• Mid (L2): L2 players who investigate pacing/mastery feel the deliberate mode rotation as the game's combat-variety discipline (Pillar 11) — the documented worldview behind why combat never repeats.
• Deep (L3): N/A at the enum level — the seven-mode taxonomy is a design/data-model instrument, invisible to the L3 convergence thesis; it does not surface or declare anything about the substrate.

Test (consistency). Verify three anchors align before population: (1) the seven proposed slugs are a 1:1 mapping onto CVD §12.2 L656-662's seven bullets with no additions/omissions; (2) HL_0049 [registries/T0_Hard_Lines .../Sheet1.csv] rule text counts over 'dominant combat type' — the R1/R2 gate must read the dominant_combat_mode column against exactly these 7 values; (3) run the topography R1/R2 gate over TOPOGRAPHY_TABLE.md's 79 rows and confirm every 'mode' value is one of the seven slugs (no orphan values) and the only violations are the already-documented R1×3 / R2×1 residuals — no NEW violations introduced by the enum itself. _Expected:_ Seven slugs map 1:1 to §12.2's seven bullets; all 79 table rows use only in-set values; gate residuals unchanged (R1×3 Ch40-41/71-72/72-73, R2×1 Ch71-75); zero orphan mode values.

Adversarial review — HOLDS. I independently re-read every load-bearing anchor and the recommendation holds on all six attack vectors.

VERIFIED ANCHORS:

• CVD §12.2 L654-666: confirmed verbatim. The seven modes at L656-662 (Direct / Stealth-restraint / Puzzle-routed / Environmental / Vril-management / Familiar-leveraged / Astral-dream-world) map 1:1 onto the proposed slugs {direct, stealth, puzzle_routed, environmental, vril_management, familiar, astral} with no additions or omissions. The R1/R2 rule sentence at L664 ("Two consecutive chapters never share dominant combat type. Every five-chapter window contains three or more distinct combat-mode dominant chapters") is exact.
• HARDLINES HL_0049 (CVD §17.10): confirmed verbatim as a canonical hard line restating R1/R2 over "dominant combat type" — binding the rule to the §12.2 seven-item list. This is the crux: the enum and the hard line it enforces share the same canon list.
• SCHEMA gap: grep for dominant_combat_mode/combat_mode returns zero matches — confirmed no pre-existing enum, so this instantiates rather than overrides canon.
• TOPOGRAPHY_TABLE.md: all 79 rows use only the seven proposed slugs (zero orphan values). The two-axis separation is REAL and load-bearing — the table carries a separate c/p/l/ch/t beat column where 'p' (puzzle beat) is independent of the 'puzzle_routed' mode (e.g. CH_05 mode=puzzle_routed / p=Y vs CH_22 mode=direct / p=N). This directly substantiates keeping puzzle_routed != puzzle. Waiver residuals confirmed: R1x3 (Ch40-41, Ch71-72, Ch72-73 — all verified against the mode column: 40/41 both puzzle_routed, 71/72/73 all direct), R2x1 (Ch71-75). No new violations introduced by the enum.
• COMBAT §9.1 L319-330: confirmed the ten categories are 'encounter categories' (texture taxonomy), and 'Social bosses... No weapon engagement' at L325 is verbatim — correctly rejected as colliding with Josh Ruling 1 and as a subordinate-doc taxonomy the R1/R2 rule does not cite.

ATTACK RESULTS:

1. Invented canon: NO — the seven modes are an existing verbatim CVD list; only slug spelling is authorial, and that is correctly routed to Josh.

2. Authority-hierarchy: NO error — Option 3 correctly rejected because COMBAT is subordinate to CVD and the rule lives in CVD/HL_0049.

3. Hard-line violation: NONE — no mode is combat-absent (environmental->§12.3 protocol, astral->§4.5 sanctuary combat), consistent with Ruling 1.

4. Stronger option: none. The rule can only be gate-checked against a closed set and CVD attaches exactly one (§12.2). The dominant-vs-combined distinction is handled correctly (enum = vocabulary; 'dominant' = the selected value).

5. Three-layer: holds — player-invisible infrastructure; L3 N/A is honest.

6. Disposition: correctly conservative. await_josh_confirm is right because slug spelling is a Josh-exclusive naming choice; apply_now would have been the error, and the recommendation avoids it.

The recommendation is the most literal possible reading of canon and survives.

Corrections to fold in:

• Minor apply-step detail: apply_actions says it will flip the taxonomy section 'from [PROPOSAL] to CONFIRMED,' but no literal '[PROPOSAL]' marker exists in the TOPOGRAPHY_TABLE.md table region (the header is a plain title). Cosmetic only — reword the apply step to 'record the CONFIRMED taxonomy ruling in the doc' rather than assume a specific marker to flip. Does not affect canon soundness.
• Recommend Josh explicitly acknowledge open_subflag #2: confirming this enum as the R1/R2 counting unit locks HL_0049's gate to exactly these seven values, so any future mode addition requires a CVD §12.2 edit. This is a real forward constraint on canon and should be an affirmative ruling, not an implicit consequence.

Apply actions. On Josh's confirm: (1) Add dominant_combat_mode to SCHEMA (_source/00_Tier_0_Master_Indices/T0_Schema_Dictionary [ACTIVE v1.3].md) as a closed enum with the seven values {direct, stealth, puzzle_routed, environmental, vril_management, familiar, astral}, each carrying its [SRC: CVD §12.2 L656-662] anchor and a one-line definition, plus a note that puzzle_routed (combat mode) is intentionally distinct from the mode_variety_check 'puzzle' beat key. (2) Populate the dominant_combat_mode column in registries/'T0_Chapter_Index [ACTIVE v1.2]'/ from TOPOGRAPHY_TABLE.md's 79 'mode' values. (3) Point the harness R1/R2 gate at that column with the seven-value allowlist; keep the documented R1×3/R2×1 residuals as the known-waiver baseline. (4) In docs/archive/TOPOGRAPHY_TABLE.md flip the taxonomy section from '[PROPOSAL]' to CONFIRMED and record the ruling. If Josh instead re-slugs, apply the same steps substituting his chosen spelling(s) everywhere and update topography_rows.json's taxonomy verdict note accordingly.

Open sub-flags for you:

• [DECISION NEEDED - JOSH] Slug spelling is a naming choice = Josh-exclusive: confirm the seven slugs verbatim, especially puzzle_routed (kept distinct from the Pillar-11 'puzzle' beat-axis key) and the underscore forms vril_management / puzzle_routed.
• [DECISION NEEDED - JOSH] Does confirming this enum as the R1/R2 counting unit also lock HL_0049's gate to exactly these seven values (i.e., no future addition of a mode without a CVD §12.2 edit)? Recommend yes — flag for explicit acknowledgement since it constrains future canon.

---

Generated by harness/site/structure_site.py — the URL path is the repo path. review root