COMPANION_CAST_BRIEF.md

npcs/COMPANION_CAST_BRIEF.md

THE COMPANION CAST — RATIFICATION BRIEF FOR JOSH

Status: RATIFIED (Josh, 2026-07-19) — see the ▶ RULINGS block below. The authoritative record is

docs/spine/DECISIONS_PENDING_JOSH.md ## D-CAST; this block mirrors it.

## ▶ RULINGS (Josh, 2026-07-19)

- ROSTER = Option B (§6): the three-tier shape ratified (core 6–8 / guests 6–10 / contacts

15–30; travelling BOND-bearing cast ≈ 12–18 — this closes the R2 roster-size question; the ~30

was the contact tier). Guest promotions stay gated on the §5 book-processing dependencies.

- CORE FOUR ratified as presented (§3.1–3.4 recommended tuples): Ntala for_life · Ana

come_and_go FULL (parts at Yggdrasil Ch 74) · the Thracian gladiator invested_sticky FULL ·

Binti Hassan invested_sticky ship-keeper (the triple-cast resolution stands).

- Ch 2–7 stays SOLO as built — the thinnest-witnessed stretch is a deliberate

arrival/assimilation choice; no early recruit, no early-Ana expansion.

- The Thracian gladiator's name = Bithus (single birth-name, no arena-name pair).

- Ruled earlier within this brief (2026-07-17): Ntala's Evil-path INVERSION — for_life never

breaks; the bond inverts into an endgame opposition encounter (design brief is the queued follow-on).

The body below is the ratified analysis, retained as the record. Every casting, name, culture, and

loyalty value was a labelled RECOMMENDATION grounded in the spine's already-staged cast + real cultural

substrate in the corpus (RCM, spine nodes, registries); the ▶ RULINGS block above states which now bind.

Findings an adversarial or coherence pass REFUTED are presented already corrected, with the

correction named. The two burned lane failure modes are held throughout: no measured number hides an

undisclosed filter, and no seed/proposal is dressed as canon.

How to read this in one pass: §1 is the whole cast on one screen. §2 is who rides when. §3 is the

per-companion detail. §4 answers your open roster-size question. §5 is the book-processing dependency

list. §6 is the roster-level options/recommendation/objection. §7 is what your ratification unblocks.

---

1. THE CAST AT A GLANCE

Ratified ground first (rowed / spine-proven), then the designed recommendations, then the correctly-rooted

locals, then the axis-separate figures listed only so the endgame crowd is not miscounted as companions.

Rows 1–2 are ratified ground (ROSTER-B). Rows 3–8 are the pass's designed recommendations. Rows 9–11 and

the two un-numbered tiers are the correctly-rooted / axis-separate figures — kept in the table so the cast

is honest about what is NOT a travelling companion.

---

2. THE ROTATION MAP — who rides Ch 2 → Epilogue (one screen)

ANCIENT ARC (Ch 2–38) — "gather the crew"
 Ch 2–7   protagonist mostly SOLO   ······ Ana intermittent from Ch 3 (watcher/vouches); Elder Rato local
          ▲ THINNEST-WITNESSED STRETCH — FOUNDATION A gap #1, only PARTIALLY closed (see §4/§6)
 Ch 8     ★ Binti signs as ship-keeper  (ship acquired on the Swahili coast; covers Ch 8–12 witness gap)
 Ch 13    ★ Ntala active join  (silent watcher since Ch 3)
 Ch 27    Ana rejoins
 Ch 28    Ana departs (→ Ch 41)  ·  ★ Thracian gladiator recruited on the free-the-ludus path
 Ch 29–40 gladiator + Ntala + Binti  (gladiator is SOLE melee — fills the measured Ch 29–40 vacuum)
 Ch 41    Ana re-travels   ·   Amboko (Ch 40) + Oupa Gert (Ch 41) = rooted light contacts, left in place

MODERN ARC (Ch 39–65) — "test the crew"
 Ch 39–42 Tunde carried lightly (LIGHT_ALLY; GUEST only if upgraded)
 Ch 44/52 Kanenhí:io / Nayra Condori — teachers, NEVER recruited (sovereignty-in-place is the theme)
 Ch 55–56 Ana's BUILT integrity-forcing WITHDRAWAL on the corrosive branch; Wenceslao covers the slot
 Ch 59–65 site-keepers = rooted LIGHT_ALLY  ·  Sōma (Ch 60) bonds, stays Japan-rooted
 Ch 65    ▓▓ THE LOCK / funnel ▓▓  Binti + gladiator + all Earth-bound allies SHED (Earth is theirs)
          Tunde + Sōma stand in the coalition = THE STANDING payoff (not aboard)
          Only Ntala + Ana carry forward

ENDGAME REALMS (Ch 66–77) — "only the truest cross"
 Ch 66–74 Realm Road: protagonist + Ntala + Ana + childhood familiar + war-host
 Ch 74    ★ YGGDRASIL CROSSING: Ana + found-family PART; Ntala + familiar cross (Ntala pays the time-price)
 Ch 75    fairy realm at war: Ntala (record) + familiar/war-host + CF-2 (grown friend) + Arawn crossed
 Ch 76    Two Courts: Ntala's journal answers Cassius; the father (Phase 2)
 Ch 77    Waterfall: protagonist ALONE at the summit + familiar (+ father on TG/GE); Ntala at the foot

Party load never overcrowds: peak simultaneous core aboard is ~4 (Ntala / Ana / gladiator / Binti) plus

rotating guests — a ship's-company shape, not an army.

---

3. PER-COMPANION CARDS

3.1 NTALA — FULL_COMPANION — the emotional spine and the House's antithesis

• Who / why she matters. Manggarai, Mama Klara's child, a non-vril human scholar. Her evidence-journal

IS the game's documented-lore system and the structural antithesis of the House's knowledge-erasure —

"everything the mine would erase, the journal holds." She is the anti-chosen-one counterweight, the

deepest assimilation-witness (she is seen while the protagonist stays hidden), and the best-friend echo:

the Ch-1 best-friend choice seeds her gender/skills/combat-focus, paid off at the Ch-75 homecoming.

• When. Silent watcher Ch 3 → active join Ch 13 → present through the Epilogue (55 of 79 nodes).
• Wiring. Loyalty for_life; two SCRIPTED integrity departures exist in the spine (an early one and the

Ch-65 EVIL lock) and must be reconciled with the "for_life is exempt" clause at the wiring commit. Trade:

the journal / parallel-intelligence (Ch 16 Indus distributional decode — documented scholarship, no

magic); RECOMMENDED secondary Trade-4 Herbalism PEER as matrilineal inheritance (flagged, kept secondary).

• Growth. Unsigned watcher (Ch 3–12) → journal-keeper (Ch 13) → the one whose testimony ends the House

(Ch 76). On the recommended endgame she crosses Yggdrasil at the price of her mortal world.

• CORRECTED (was refuted). An earlier draft gave her a combat verb MARK/EXPOSE operating inside boss

Composure combat, plus a Herbalism "combat facet." This contradicts the supreme character authority:

flagship T3_Core_Characters §3.4 locks the Healer's Child as "No combat abilities — informational and

social capability only… sharper than the protagonist's vril-based intuition **in strictly human social

contexts." A boss Composure node is a combat context, so the verb violated the lock. Corrected

position: Ntala's contribution is a non-combat recon / social-read** (she reads NPC behavioral cues,

exposes a lie or a forgery, marks a target for the party OUT of the combat grammar) — consistent with

"informational and social capability only." Whether the general "allies wield combat too" mandate should

override the flagship §3.4 no-combat lock for her specifically is [CARDINAL — characterization] for

Josh, not a thing this design settles.

• Cardinal items. belief_axis token (proposed truth_witness_mercy); the Herbalism strand (rec vs

journal-only); reconcile the two for_life departures with the exemption clause; culture is Manggarai

per flagship T3_Core (supersedes the older "soft African-marsh" read — a CH_13 body-prose wrinkle to

smooth); CF-2 reunion-as-distinct-NPC (pre-existing open cardinal). Name Ntala is RATIFIED (2026-07-12).

3.2 ANA — FULL_COMPANION — the friend who leaves and keeps choosing to come back

• Who / why she matters. Manggarai caci stick-fighter, the freely-chosen peer opposite Ntala's fated

echo. Load-bearing to five ratified systems at once: the sole BUILT exemplar of companion-loyalty-IS-

integrity (her Ch 55-56 corrosive-forcing withdrawal, covered by Wenceslao so no wall); the party's only

pure-physics / non-vril combat anchor with DESIGNED vulnerability (breaks the DIRECT Composure node with

body-grammar, guards the body during astral-projection at Ch 44); the House's chosen human lever (the mara

set on her, the Ch-61 wound); the assimilation witness who STAYS ("my mother would have known what you

are"); and one of the few who cross the Realm Road toward the fairy-realm homecoming.

• Why come_and_go IS her meaning. A for-life companion cannot give the bond ladder its return-texture —

you cannot feel a homecoming from someone who never left. Her three-sentence-max voice ceiling makes every

word land; the Ch-41 return line is her largest emotional expenditure. Verdict FULL, not GUEST:

come_and_go is her rhythm, not a ceiling on her weight.

• Wiring. come_and_go; keep the live major_opposition departure (trigger 60) over the WS7-plan

circumstance_only, because her CL-02 withdrawal is integrity-driven and BUILT.

• Cardinal items. FULL vs GUEST label (rec FULL); belief_axis honor_and_freeing_the_bound (grounds her

caci fair-contest ethic); restore the T3 wariness seed (resentment → competitive fascination) with the

Ch-3 vouching arriving AFTER she reads the protagonist's caci skill; **does Ana cross the Ch-65 lock into

the realms, and by what justification** — she is staged present through Ch 74–75 yet is mortal, non-vril,

no birthright (a genuine world+characterization fork; the recommended endgame parts her at Yggdrasil Ch 74);

Manggarai-authenticity screen on the name "Ana" (low priority).

3.3 THE THRACIAN GLADIATOR — FULL_COMPANION (invested_sticky) — the liberation spine given a face

• Who / why he matters. An ancient Thracian, enslaved young, freed in the Ch-28 ludus breakout. Three

measured facts converge (this is grounding, not invention): he fills the arc's largest measured melee

vacuum (Ch 29–40, the exact chapters Ana vacates); he gives the already-canon-but-FACELESS liberation

lineage (ER-2 captive→competitor→liberator; today a nameless cohort) its person; and he is a first

exemplar of the ratified-but-empty invested_sticky model, on a DIFFERENT belief pole from Ana (property/

bondage vs her forcing-seizure axis — he does not reassign her ratified instance).

• When. Recruited Ch 28 on the free-the-ludus path (non-gating: other paths leave the slot empty);

present-continuous Ch 29 → 65; a 33-chapter homecoming at Ch 61 (Bulgaria/Thrace — he stands beside the

Thracian-heritage scholar Kamen Iliev, defending his own people's survived memory against his old

enslaver). Does NOT cross into the realms — and that is the payoff: for the first time he is free to

choose his road, and chooses to lead the freed on Earth.

• Wiring. invested_sticky (high inertia, but the 90-cap invariant holds — complete opposition still

forces departure); belief axis freedom_from_bondage; departure major_opposition (60); combat = heavy

shield-and-*sica* GUARD-break, real-physics, no vril; crew station gunner / master-at-arms + training-yard;

not a trade teacher (combat is not one of the 10 trades — inflating him would be padding). Richest hook:

he keeps the protagonist's secret because he has kept worse, and LEAVES when she becomes an owner.

• Cardinal items. Casting a new named companion (mint CHAR_0NNN); the NAME (rec Bithus — attested

Thracian, appears as a gladiator's name in Horace; plus an optional arena-name-vs-birth-name pair);

culture = ancient Thracian; gender (rec male, for combat contrast with Ana; gladiatrix attestable);

FULL vs GUEST (rec FULL); the Ch-61 homecoming as additive; the Ch-65 disposition + any Ch-75 coalition

role.

3.4 BINTI HASSAN — FULL_COMPANION (invested_sticky ship-keeper) — the ship's soul

CORRECTED — was triple-cast. Three parallel lanes designed CHAR_0056 into two incompatible roles: a

come_and_go GUEST (sea-legs, Ch 8–38 Trade-10 teacher, Azores reunion Ch 38) AND an invested_sticky FULL

ship-keeper aboard Ch 8–65. A single character cannot be both. **Resolution (coherence critic's synthesis,

presented as the recommendation): make Binti the invested_sticky ship-keeper**, and FOLD the come_and_go

"sea-legs guest" material into her pre-signing Ch-8 recruitment texture. This collapses two designs into one

richer character, fills the Ch 8–12 witness gap more continuously than discrete crossings would, and removes

the Trade-10/Swahili/mtepe duplication. Final casting is Josh's.

• Who / why she matters. Swahili-coast dhow captain; the ship's OWN culture — the flagship VH_001

descends from her mtepe, acquired Ch 8 — so her helm/navigator station is native, not bolted on. She is

the witness the field companions can't be: not the witness to each CULTURE but to the CREW and to who the

protagonist becomes BETWEEN cultures — the confidant you come home to, changed. Her belief pole is the

moral inverse of the House's 5,000-year extraction fleet ("no ship serves bondage").

• When. Join Ch 8 (first crew-companion, five chapters before Ntala) → present Ch 8 → 65 → shed at the

lock (the mothership cannot enter the realms); the hearth of the homecoming Epilogue.

• Wiring. invested_sticky; belief axis free_passage_no_ship_serves_bondage; departure

major_opposition (60), crucible = the Ch 33–35 slave-trade waters (the C7 care-firewall becomes a

character stake); Trade-10 Fishing & Navigation teacher + Trade-8/9 hull-reconfiguration peer; combat verb

= the SHIP (silent-running, vril-burst keel — a naval/environmental Composure node, no melee duplication);

Tier-3 bonded attackability.

• Cardinal items. The triple-cast resolution above (invested_sticky keeper vs come_and_go guest — the

sharpest single call in the roster); belief-axis token; a promote-to-CORE option (for_life Trade-10

navigator holding the mothership helm, given the VH_001-from-her-dhow lineage) if Josh wants a second

for_life; the Vehicle_Registry drift (VH_001 rowed air / CH_38 vs the doc/contract's sea-hull-at-Ch-8) —

flag, do not resolve; ELEVATED-care Swahili maritime substrate is a book-processing dependency (§5).

3.5 TUNDE — LIGHT_ALLY (GUEST on upgrade) — the enemy runs on people it bought, and each can be un-bought

CORRECTED — verdict lean inverted. The prior draft LED with GUEST_COMPANION as "the recommendation."

Adversarial + coherence both find the honest present-state is LIGHT_ALLY: the authored spine carries

Tunde "lightly" in the exact same breath as Amboko and Oupa Gert (Tier-3 contacts, no loyalty row). GUEST

is a real promotion, not the floor — and it is gated on book-processing the modern-Nigeria substrate. So:

default LIGHT_ALLY; GUEST = explicit upgrade conditional on (a) Josh wanting the drawn_conditional

exemplar authored AND (b) the modern-Nigeria enrichment.

• Who / why he matters. Modern Lagos; the strongest single body in all 162 rows for the zero-exemplar

drawn_conditional model, because his authored Ch-39 fork already IS the drawn-by-integrity idea:

generative "free-and-turn" draws him in, corrosive "use-and-discard" spends him. He is the fastest-leaving

bond (the canary): he leaves when your integrity resembles the masters he broke from.

• When. Ch 39–42 (carried lightly); recurs at Ch 75 through his network — a STANDING-axis coalition

payoff, which is why guest and light-ally pay off IDENTICALLY there (part of why the promotion is optional).

• CORRECTED cost + hook claims. The prior "promotion needs NO new chapters — a 3-node re-flesh" is

softened: converting fold-name baggage in three LOCKED chapters into a present, voiced actor IS real

authoring against the locked 77, and the persona interior is book-processing-blocked — so it *reduces, does

not remove,* the cost, and is itself gated on processing. The "assimilation-witness from the inside" hook

(his NDA concealment-under-duress → he recognizes the protagonist's hiding) is a **design hypothesis,

marked post-processing-to-validate**, not a present corpus-grounded asset.

• Wiring (if upgraded). drawn_conditional; belief axis freedom_from_bondage_redemption; departure

any_sustained_misalignment (35, the canary); combat = stealth/puzzle-routed access+decode, ordinary

human, no vril; Trade-9 Inscription modern peer; crew station fixer/quartermaster.

• Cardinal items. LIGHT_ALLY vs GUEST (rec LIGHT floor, GUEST as gated upgrade); ratify or retire

drawn_conditional (Tunde is the primary candidate exemplar); belief-axis token; the departure threshold

(35 vs the WS7-plan major_opposition); name Tunde/Babatunde (registry-finalized 2026-07-10, light flag).

3.6 SŌMA KŌHEI — GUEST_COMPANION (optional 2nd drawn_conditional) — the record that won't be filed

• Who / why he matters. Japanese master swordsmith (CHAR_0154), the fullest-staged Tier-3 figure (6

actor-beats). She forges a blade beside him over three days — a shared-labor bond, not a rescue — and he

teaches the arc's moral core ("the hardest part of any blade is knowing when not to draw it"). He is the

Ch-60 delegated-victory: an ally standing the record in daylight collapses the Factor's concession.

• When / §17.1-safe shape. Ch 60 is built on an antagonist who EXTRACTS Japan's heritage OUT of Japan —

so lifting a swordsmith out of Japan would rhyme with the villain. He therefore stays **Japan-rooted and is

DRAWN to stand at the Ch-65 Gathering** (summoned witness, not extracted asset) on a sustained-generative

run. GROUNDED substrate (tatara/kobuse/hamon/NBTHK, deep-research-verified) — no fakery.

• Wiring. drawn_conditional; belief axis record_kept_in_daylight; departure major_opposition (60);

combat = the RESTRAIN verb (cut-and-sheathe); Trade-3 Blacksmithing teacher; forge/armory station as a

summoned/away contributor, not live-aboard.

• Cardinal items. Whether ANY modern figure travels at all; Option A (promote Sōma) vs Option B (author

a new modern-professional, e.g. a restitution lawyer with no §17.1 rootedness problem); the loyalty tuple;

the belief-axis token; the name re-opens at NAMING_METHOD visibility on promotion.

3.7 CF-2 — THE GROWN BEST FRIEND — GUEST_COMPANION — the first opening choice paying off at the last

• Who / why she matters. Fairy-realm / elven, realm-native. The one she grew up beside, left behind in

the realm when she fell through the seam, alive and grown as one of the realm's own. The Ch-1 Selection-2

choosing paid off at the Ch-75 homecoming; the designer-only resonance closes here (Selection-2 SEEDED

Ntala, so "the friend she grew up beside and the companion who began as her mirror" stand in one moment —

felt, never spoken). Reveal-orthogonal (carries NO family who-frame / thesis / Grand-Sage content).

• When. Ch 75–77, rooted in the realm — canon already ruled her a NON-travel companion (the old mid-game

travel-companion idea is dead). Her value is emotional closure, not party mechanics.

• Wiring. If she takes up arms in the coalition, drawn_conditional is the only fitting token (she never

travelled, so the model enum barely applies) and she is a standing candidate to give that model its

endgame-variant exemplar. Substrate is GROUNDED game-canon — Josh-authored, NOT book-processing (the

fairy realm cannot be over-researched).

• Cardinal items. Her NAME (naming-canon); GUEST (takes up arms) vs LIGHT_ALLY (pure reunion cameo);

the CF-2 reunion-as-distinct-NPC call (shared with Ntala's card).

3.8 ARAWN — GUEST_COMPANION (war-guest) — the standing that pays out

• Who / why he matters. Welsh, sovereign of Annwn (the Otherworld of the Mabinogi, Ch 73) — a LIVING

tradition, ELEVATED §17.1 care, GROUNDED in a full deep-research pass (Davies, Parker, Haycock, Koch).

Honored, never fought; "by the end, a firm friend." He crosses in person at Ch 75 to answer a welcome he

still owed — a THE STANDING (faction-reputation) payout, not a THE BOND recruitment.

• Wiring. Not a loyalty_model row — he is the demonstration that a door left open owing/owed sends aid.

Care-firewalled: a distinct honored Welsh Otherworld, never conflated with the fairy realm.

• Cardinal items. Confirm named war-guest ally status and the honored-guardian aid model (STANDING, not

recruitment).

3.9 THE ROOTED LOCALS — Oupa Gert, Amboko, Wenceslao, and the sovereign keepers — LIGHT_ALLY / NOT_A_COMPANION

• The split is the finding, and rooted is the RIGHT answer, not a demotion. These carry per-culture

authenticity at coverage scale and are the honest home of the assimilation mechanic — the local you earn

and LEAVE IN PLACE.

• Oupa Gert (‡Khomani, Ch 41) — richest MEANING, LEAST promotable: his written arc refuses being carried

off ("left with his grandchildren, not carried off in her book"); §17.1 elevated-care inviolate.

• Amboko (Mbuti, Ch 40) — book-processing-BLOCKED past Turnbull + a witnessed-only religion; travelling

him would fake depth (a §17.1 violation).

• Wenceslao (Lickanantay, Ch 55–56) — the one modern rooted ally with a pre-seeded guest hook (the

unbuilt Ana-slot cover). LIGHT_ALLY with a flagged guest-upgrade path, gated on a sovereignty-protecting

travel justification.

• Kanenhí:io (0134), Nayra Condori (0147), the 59–65 site-keepers — protected sovereign keepers whose

whole narrative function is sovereignty-in-place. Promoting any to a traveller negates the beat and breaks

§17.1. NEVER travel; the honest path to a recurring guest from any of them is a book-processing pass FIRST.

3.10 THE AXIS-SEPARATE FIGURES — listed so the endgame crowd is not miscounted

• The childhood familiar — the 22-slot FAMILIAR axis (lost Ch 2, reunited Ch 58, crosses home Ch 75,

leads the war-host menagerie). No loyalty_model, not on THE BOND. The party's heart, but never a companion.

• The Elven King / father (CHAR_0005) — endgame reveal figure, name-forbidden (HL_0099); a path-

conditional Waterfall guest, never a party member.

---

4. THE NUMBER — recommended total vs the unratified ~30

The ~30 is an ESTIMATE, single-sourced, not canon. It occurs in exactly one authored line

(WS7_WIRING_BUILD_PLAN L158, "do NOT mass-populate the ~30 companion rows before Josh ratifies"); every

other appearance cites back to it. It is not in CVD, any T1, or the persona doc. So the roster question is

not "which 30" — it is that the target itself was never ratified.

What the spine actually proves: two long-haul travellers (Ntala 191 beats/55 nodes, Ana 44/22), plus a

thin tier of single-region light contacts. It does not gesture at ~30 travellers. The "~30" almost

certainly counted the Tier-3 local-contact roster — the identical misread this lane already caught on

Tunde (a rowed light ally mistaken for a cast member).

The recommended answer is a SHAPE, not a single headcount (a One-Piece core + ME2 loyalty arcs + FTL

station grammar, funnelled by the Ch-65 lock — explicitly not a BG2 roster or a Suikoden army):

Travelling BOND-bearing companion cast (Tiers 1+2) ≈ 12–18. Endgame-realm party ≈ 2–4 + the familiar.

The coherence critic's reasoning for why this shape is sound: before this pass only 2 of the 5 ratified

loyalty models had exemplars (for_life = Ntala, come_and_go = Ana). The new casts now exercise ALL FIVE —

invested_sticky (the gladiator AND Binti), drawn_conditional (Tunde / Sōma / CF-2), not_companion

(contacts + familiar tier) — and the departure-capable bonds that HURT fire on DISTINCT belief axes (Ana's

built Ch 55-56 withdrawal; the gladiator's property/bondage axis; Binti's slave-waters axis; Tunde's

canary-35). The protagonist's three ratified hooks each have a clean owner: the best-friend seed (Ntala,

paid off by CF-2), the hiding (a differentiated assimilation-witness spectrum), and the homecoming.

Two coordination artifacts the critic flagged, corrected here:

• invested_sticky would have TWO exemplars, not one. Both the gladiator's card and Binti's card

independently claimed to be "the first invested_sticky exemplar." Two is fine and desirable (different

belief poles) — Josh is filling the zero-exemplar slot ONCE, with two characters, not being told twice.

• drawn_conditional is over-provisioned — sequence it. Tunde (Ch 39), Sōma (Ch 60), and CF-2 (Ch 75)

each claimed to be the sole filler. Recommended sequence: Tunde = primary/ratified exemplar (modern

arc, existing seed, name finalized), Sōma = optional second, CF-2 = the endgame reunion variant —

not three competing "the one."

---

5. THE BOOK-PROCESSING DEPENDENCY LIST

Where a culture's corpus substrate is THIN, the persona is flagged book-processing-blocked — never faked

(§17.1: care = weight + authenticity). Josh's own instinct — strengthen during quest-writing / book

processing — is the right sequencing. Finalize these personas AFTER the enrichment pass, not before.

NOT book-processing (do not mis-flag): Sōma Kōhei (GROUNDED — Ch 60 is richly built), CF-2 and the

whole endgame (GROUNDED game-canon — the fairy realm is Josh-authored, cannot be over-researched), Arawn

(GROUNDED — the Ch-73 Welsh deep-research pass). The endgame gaps are cardinal-canon (names, the crossing

ruling, reveal solitude), the exact inverse of the Earth-arc thin-lane risk.

---

6. ROSTER-LEVEL OPTIONS + RECOMMENDATION + STRONGEST OBJECTION

The decision: how much of this roster to ratify and row now.

• Option A — ratify the full designed roster now (all of §1 rows 1–8 with recommended loyalty tuples).

Pro: gives the WS_027 wiring commit a rich fixture; one pass. Con: several personas are book-processing-

blocked (§5), so ratifying their interiors now would either freeze thin values or invite the do-not-invent

breach; mass-population is exactly what WS7 L158 forbids pre-ratification.

• Option B (RECOMMENDED) — ratify the SHAPE + the spine-proven core; stage the rest as gated upgrades.

Row Ntala + Ana (proven) and the two clean core recommendations that are structurally load-bearing (the

gladiator, Binti-as-ship-keeper) with recommended tuples; carry Tunde / Sōma / CF-2 / Arawn / Wenceslao as

a labelled guest tier whose promotion is gated on the §5 dependencies; ratify the three-tier size shape

(§4) as the answer to the ~30. Pro: every rowed entity is spine-proven or structurally forced, zero

invention; unblocks the wiring; turns the open question into a bounded ruling. Con: leaves the guest

interiors formally open — but that is the honest state, and it is what book-processing is for.

• Option C — status quo (row nobody beyond the two; defer all design to the wiring commit). Pro: maximal

do-not-invent purity. Con: re-orphans companion_roster_state (the recurring anti-orphan failure); throws

away real, adversarially-verified design evidence to avoid a decision.

RECOMMENDATION: Option B. It rows only what is proven or structurally load-bearing, honors do-not-invent

and the two burned failure modes, unblocks the wiring with a real fixture, and hands you a bounded

ratification (a shape + a core + a gated guest tier) instead of a bare "which 30?"

STRONGEST OBJECTION: Option B quietly ratifies "the travelling cast is ~12–18, not ~30," and the ~30 may

be a real authorial intent the living spine simply has not staged yet (WS1/WS3 enrichment could still cast

more homeland recruits). Why B still wins: it does not foreclose the number — it makes the number a

DECISION. If you rule the intended core is larger, the consequence is a concrete authoring work-order ("the

spine casts 3 core + Ana; cast N more homeland recruits, ~1 per 4–6 ancient chapters"), NOT a registry

mass-population — the do-not-invent-compliant path. And the honest residual weakness the critic named

stands: the ancient first half (Ch 2–7) is still the thinnest-witnessed stretch — Binti's Ch-8 signing

closes Ch 8–12, but Ch 2–7 has no aboard witness (there is no ship yet). If you want that stretch covered,

it is a deliberate early-recruit or an early-Ana-presence call, not something this roster silently fixes.

---

7. WHAT RATIFICATION UNBLOCKS

Once you rule §4 (the shape/size), §6 (Option B), and the per-card cardinals:

• Registry wiring (WS_027 / D-REL-BANDS). Row the ratified core with loyalty_model / loyalty_score

(−100..+100) / belief_axis / departure_threshold / join_trigger, and bind THE BOND ladder end-to-end

against a real fixture (Ntala for_life, Ana come_and_go, the gladiator + Binti invested_sticky). Refresh

docs/fidelity_baseline.json in the same commit (the gate goes red otherwise). De-orphans

companion_roster_state.

• Quest hooks. The gladiator's Ch-28 recruitment fork + Ch-61 homecoming beat; Binti's Ch-8 signing +

Ch 33–35 slave-waters crucible; Ana's Ch 55–56 withdrawal wiring + Wenceslao cover; Tunde's Ch-39 draw

fork + Ch-75 coalition; Sōma's Ch-65 muster; the Yggdrasil crossing (Ch 74) + reveal-solitude trim (Ch 77).

• Naming-method runs (docs/NAMING_METHOD.md). The Thracian gladiator (Bithus + arena/birth-name pair);

CF-2's name; belief-axis vocabulary tokens (truth_witness_mercy, honor_and_freeing_the_bound,

freedom_from_bondage, free_passage_no_ship_serves_bondage, freedom_from_bondage_redemption,

record_kept_in_daylight); the Manggarai screen on "Ana"; the Sōma name re-open on promotion.

• Book-processing queue (§5). Schedule the Thrace, Swahili-maritime (ELEVATED), and modern-Nigeria

passes before those personas finalize — the Tunde GUEST upgrade is explicitly gated on the last.

• Reconciliations to close at the wiring commit. Ntala's two for_life departures vs the exemption

clause; Ntala's non-combat recon framing vs the §3.4 flagship lock; the Binti triple-cast resolution; the

Vehicle_Registry VH_001 air/Ch-38 vs sea-hull/Ch-8 drift.

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