decisions/spine/NTALA_INVERSION_ENCOUNTER_BRIEF.md
Status: PARTIALLY RULED (Josh, 2026-07-19) — D-1 SEATING = A + the Ch-65 seed; D-5 TRANSIT =
crosses behind the record; the D-3/D-4/D-8 riders rode without objection. **D-2 (the §3.4 lock) is
NOT ruled — Josh CHALLENGED the keep-the-lock recommendation** ("She will be a powerful foe... she has
to stop you") and ordered a full fight-design deep-dive. **That brief has LANDED and SUPERSEDES this
doc's D-2 branch analysis: docs/spine/NTALA_D2_POWERFUL_FOE_BRIEF.md** (recommendation F-★ "The Read
Made Kinetic"; alternative F-C "The Peer Blade"). Authoritative record: D-NTALA-ENC in
docs/spine/DECISIONS_PENDING_JOSH.md.
Beat-id note (2026-07-27): the beat this brief calls CH75_B04a was renumbered to CH75_B13 in
docs/spine/CH_75.md (schema rule v — the letter-suffixed form was invisible to the spine parser);
the live id is CH75_B13.
Provenance: workflow wf_6bc80290-668 (2026-07-19, 8 Opus agents): canon-grounding → three
independent seating designs (A/B/C) → three adversarial lenses (canon-consistency ·
care-register · AAA-quality; reviewer rankings A>B>C and B>A>C, both agree C fails hardest) →
synthesis. Every load-bearing citation was verified against the repo by the reviewers.
Chapter assignments and any combat-ability grant are RECOMMENDATIONS only — marked [JOSH];
the do-not-invent rule binds.
---
Per Josh's ruling (2026-07-17):
for_life, never breaks. On the EVIL path the bond does not sever — it inverts: she leaves the player's *side* and becomes an additional endgame OPPOSITION encounter, the arc's moral anchor standing against what the player has become.CH_65.md #146 FORK_CH65_A EVIL row, "Ntala departs (companions who oppose the corrosion leave)" — is *superseded in spirit* by the ruling; the good/neutral rows (#144–145) keep her, untouched.ws.path_bucket: EVIL, ws.path_lock_in_state: LOCKED). No seating may offer a redemption fork; that would break the lock.What is NOT yet ruled — and is exactly what this brief asks Josh to rule — is the encounter design: its seating, whether the §3.4 no-combat lock holds, its spare/kill semantics, and its care-register commitments.
The hard constraints the design must satisfy (from the ruling, restated):
1. T3 §3.4 lock — T3_Core_Characters [ACTIVE v1.0].md:199, verbatim: "No combat abilities — informational and social capability only. Their perception of NPC behavioral cues is sharper than the protagonist's vril-based intuition in strictly human social contexts." An open cardinal asks whether Josh overrides this for her. The design must cover both branches.
2. Join the Ch-75 decision-web, do not break it — the ruled FORK_CH75_A (CH_75.md:160–167) is a 6-row terminal stand: integrity band × Ch-72 spare/kill (F1 Vritra), steadied/thinned by ws.familiar_bond_state (#156) as a modifier, not a row-multiplier. Any Ntala seating must not explode this web to a third axis.
3. Care register: grief-never-triumph — defeating her must never read as victory; spare/kill semantics per the F1 Vritra precedent (CH_72.md:138–148).
4. Reveal discipline — nothing she says may leak the who-frame / thesis pre-release windows.
5. do-not-invent — chapter assignments and numerical canon are Josh's; the brief RECOMMENDS seatings, never asserts them.
6. Rated-R 17+ mature AAA — weight and consequence, never defanged; but Ntala inverted is the game's saddest fight, not a monster.
---
CH_65.md:252, itself audit-pending D-23). The fairy realm (Ch 75–77) is a dimensional elven otherworld reached only across the Yggdrasil / World-Tree seam by the protagonist as its returning elven native (CH_75.md:7,18,19); human companions cross only on the *surviving* paths. Seating her inside the realm on the one path she is gone requires a crossing mechanism that does not yet exist in canon. All three seatings are past the threshold, so this is shared, not unique to any one — the ruling created it. It must be resolved, not hand-waved. (The most grounded option — she crosses *behind the record* at the Ch-65 threshold — is itself a reason to dramatize the Ch-65 seed, see the recommendation; it still needs Josh's ruling because a human crossing Yggdrasil without the protagonist is unestablished.)T3_Core_Characters:691). That kit fabricates a toolset she does not possess and must be dropped or re-grounded. (ii) Even the "anticipation / pattern-denial" half over-extends §3.4: line 199 bounds her superior read to "strictly human social contexts," explicitly contrasted with the protagonist's vril-based intuition — reading the player's *combat* move-chains is the vril domain she is carved *out* of. If Josh overrides §3.4, a combat capability is being invented, not unlocked, and even a granted override should derive denial from social/behavioral tells (tempo, habitual cruelty, the tells she has read for 70+ chapters), never literal combo-memorization.line 203 — "Possible departure at Ch 23 Greece if integrity strongly negative" — in tension with the ruled "for_life / inverts / never breaks," and CH_65 marks her endgame resolution audit-pending (D-23). All three seatings build on a baseline the source doc has not fully ratified. Resolve the stale line + D-23 before or with this ratification.CH_EPILOGUE.md:33: "It could be anyone… held open with care"); canon flags "resolving the unseen reader" as an at-risk FAILURE (#169). All three designs lean on "whose hand" as subtext; the reviews rank the risk B (hardest) > C > A (most disciplined). Commitment for the brief: hold ws.journal_status: NEVER_COMPLETED / ws.journal_voice_texture: SPARSE_COLD band-fixed on both spare/kill outcomes, and assign the reader NO identity, not even as subtext.Branch (i) — Lock RESPECTED (the witness-gauntlet / "unarmed line"). Recommended default.
She has no HP bar and never raises a hand. She stands unarmed in the path, the record in hand. The war's roar drops to a pocket of quiet — a still point inside PEAK intensity. Using her canonical faculty (§3.4:199, social/behavioral read), she narrates the player's *own* habits and choices — the small cruelties, the promises broken across the road — never the family, never the pattern, never the record's content. The threat is environmental: the player's own corrupted war-host is already closing on her; on Design B's variant the realm's court-guard and any found-family remnant interpose, supplying the L1 combat the lock forbids *her*. The player's verb is turn the host aside (spare) or let / order it cut her down (kill). Felling her takes no skill — she is unarmed — so it is an execution, not an encounter, which is precisely why it delivers no combat-triumph. This is the purest structural realization of grief-never-triumph.
Branch (ii) — Lock OVERRIDDEN (only if Josh grants it — invented capability, see D-6).
A "one enemy who has you memorized" fight: anticipation / pattern-denial keyed to her long social read of the player (the more autopilot the playstyle, the harder she stalls it; improvisation breaks the stall — fits the rewards-intelligence pillar). Corrections that MUST apply if granted: drop the "inverted-healing" kit entirely (she is not a healer); derive denial from social/behavioral tells, not combat-combo memorization; she fights only defensively, cannot win, only wear the player into relenting; reduce-to-standing → the same spare/kill choice as (i). The reviews caution this converts the saddest *fight-that-isn't* into *a fight*, and a defeated opponent reads closer to "boss down" than "the person who knew you, gone."
Per CH_72.md:138–148: offered never forced; both outcomes play through; no fail-wall; care_band elevated on both (ELEVATED care carrying the maximum-register duties — RULED elevated, seventeenth sitting 2026-07-28); the weight rides the manner, not the choice; neither branch re-coded good/evil. Precision the reviews add: F1 Vritra is *path-invariant*; the Ntala fork is *EVIL-band-gated*. So the spare/kill is not "the good option" — it is texture within the fall, the grave versus the gentler damnation.
ws.ntala_inversion_state: STANDING_WITNESS.ws.ntala_inversion_state: SILENCED.journal_status: NEVER_COMPLETED / journal_voice_texture: SPARSE_COLD — the mechanical signature of the inversion and the exact state the fixed Evil Epilogue renders (CH_EPILOGUE.md:94,143,214). Neither branch moves journal_status.---
FORK_CH75_D at CH75_B04a)*Corrected citation:* the FORK_CH75_C EVIL row is ws.companion_roster_state: coalition_sparse (CH_75.md:191) — the "markedly more alone" phrasing the design borrowed is actually the CH_65:146 lock cell; minor slip, not load-bearing. Golem-exclusion correctly handled (Ntala wired away from BE_0075.P3, the colossus with no hidden director, CH_75.md:141).
Seat a new EVIL-only beat immediately after the coalition muster (FORK_CH75_C), gated ws.path_bucket == EVIL. It fires into the *negative space* of the corrupted coalition: where the good path fills with allies-who-came-for-her, the EVIL path's answer to "who is left" is the ally who left, now standing in the path. FORK_CH75_A stays 6 rows, untouched; the new fork adds 0 rows to the four non-EVIL quadrants and is a 2-row sub-fork (spared / felled) inside the EVIL band. Her spare/kill runs parallel to, never crossed with, the Ch-72 Vritra spare/kill.
CH_75.md:235). Ntala *is* the record of the family pattern; giving her a speaking role here sits one slipped clause from leaking the who-frame. Mitigation (self-flagged, reviewers call it "high-wire but threadable"): confine her strictly to the player's own habits, zero agent-noun, "keeper of your soul, not of the family's crime," with a mandated grep/critic gate on the beat's body prose before commit. Note her *presence* does not leak — the House is player-facing (hostile recognition) since ~Ch 38; only her *dialogue* could. Secondary con: a quiet unarmed still-point risks being tonally swallowed by the PEAK Last-Coil / colossus spectacle (CH_75.md:148). Tertiary: a single beat gives the inversion less *build* — it collapses the ruling's "beginning" and payoff into one chapter (addressed by the recommendation's Ch-65 graft).FORK_CH76_C at the B02→B03 seam)On surviving paths the journal is the weapon at CH76_B03 (CH_76.md:93–96, THREAD_20 ANCHOR, object:journal) that answers Cassius; on EVIL that beat does not fire — the player becomes Cassius (VICTORY_EVIL, #148). B fills that exact void: the same journal that answers Cassius on the good path is, on EVIL, read against the player, in the father's hearing. The good-path weapon becomes the evil-path indictment — the only design that hijacks a specific built beat rather than inventing a free-standing obstacle.
CH_76.md:83–87); she testifies to the player's deeds, never the thesis / Grand Sage / who-frame (those stay held to Ch 77). The safest of the three on the hardest constraint. Maximal dramatic economy: inverts a real built function into the exact dramatic void EVIL leaves — no invented obstacle. Is the felt *cause* of the Epilogue's only path-variant (journal completeness). The crowding worry largely self-neutralizes because the rebuttal doesn't fire on EVIL.COMBAT_EVIL death (CH_76.md:160), displacing his canonical own-agency and 10,000-year foreknowledge (HL_0116, #14,127) and corrupting *grief-as-defense-never-villainy* (#122). B's mitigation — her record confirms what the King already senses through his own grief-made door, never informs him — is grounded but the care reviewer calls it thin, "precisely because she literally recites the ledger in his hearing at that seam." Requires a hard authored wall and ws.father_encounter_variant left on its Ch-65 lock. Secondary: misses constraint (2) — relocates the inversion out of the Ch-75 web entirely, into the game's most load-bearing reveal room (the double-climax hard-line node), competing with Cassius for oxygen. Asserts the Epilogue-reader identity hardest of the three (must be walked back per D-8).FORK_CH77_FOOT)Decouples the bond at the Ch-65 ice (dramatizing the ruling's "inversion BEGINNING, not a departure" — replacing the bare #146 cell with a staged inversion-*seed* at SC_CH65_threshold_crossing / CH65_B10), then follows the record to a terminal encounter at the foot of the mirror-waterfall in Ch 77, so the last human face before the solitary climb is the one witness who cannot be deceived — rhyming with the pool that "gives back the protagonist's face" (CH_77.md:37).
combat=N, intensity STANDARD (the resolution-exhale after the Ch 75/76 peaks) (CH_77.md:59,85,87,89,207; the one sanctioned lift authored NEUTRAL, #3,11,12). Inserting an EVIL-only opposition fork carrying a spare/kill (lock-respecting) or a boss fight (lock-override) into the one chapter whose foundational premise is the *refusal* of combat and failure is a structural contradiction with the chapter's own character — unlike A's reveal risk (a discipline problem a gate catches) or the shared transit gap (a mechanism can bridge), no authoring discipline reconciles it.reflective_pool — "where the fall gathers and lies mirror-quiet… gives back the protagonist's face" (CH_77.md:76), the summit site of the single sanctioned reveal (B05–B08). There is no separate foot-pool in canon; the climb rises *past the father's hall* up *to* the fall (#25). C risks placing Ntala at the most protected beat in the game.#33). Pre-loading it with a fresh "you just killed / passed the witness who knew you" risks *captioning* the uncaptioned, player-completed recognition — the highest-value protected moment in the arc.CH_65:146 cell.---
Ratify Seating A (lock-RESPECTING "unarmed line" at Ch-75, FORK_CH75_D), grafted onto Seating C's Ch-65 dramatized inversion-seed as the ruled "beginning," and KEEP the §3.4 no-combat lock. [JOSH — cardinal on the seating chapters and the lock]
This is the synthesis both full reviews converge toward from opposite lenses (the care review recommends it explicitly; the canon review ranks A a close second and endorses lifting C's Ch-65 seed). It is the only configuration that satisfies all of the ruling's hard constraints at once:
FORK_CH75_A at 6 rows (0 quadrant multiplication; 2 EVIL-only rows). B and C both relocate the inversion *out* of the web.Ship the lock-respecting branch as primary and hold the lock-override branch as the ratification alternative for Josh's §3.4 call (D-2) — and if he grants it, apply the D-6 corrections (no healing kit; social-derived, not combo-derived).
**Seating B is reveal-safe *by construction*; Seating A is not.** The single strongest case against the recommendation — and the reason the canon-consistency review ranks B first — is that A stakes the game's most reveal-charged character (Ntala *is* the living record of the family pattern) at the game's most reveal-locked beat (Ch 75, "no cabal-noun anywhere, never even invoke-and-negate," CH_75.md:235). One slipped clause leaks the who-frame at the worst possible window. B eliminates that risk *categorically* by seating her *after* Cassius is named — she can testify freely to deeds — and does so with more dramatic economy (inverting a real built beat, the journal-rebuttal, into the exact void EVIL leaves) than A's free-standing pocket. Accepting A means accepting a high-wire, discipline-dependent reveal risk when B offers a construction that *cannot* leak.
Why the recommendation still holds against it: (a) A's reveal risk is bounded and *gateable* — a discipline problem a mandated grep/critic pass catches — whereas B's central risk (making Ntala causally responsible for the father's grief-war, corrupting the arc's unimpeachable moral north) is *structural*, and the care review judges B's "confirms-not-informs" wall thin because she literally recites the ledger in the King's hearing at that seam; (b) B fails constraint (2) — it relocates the inversion out of the Ch-75 web the ruling requires it to join; (c) B crowds the double-climax hard-line node and competes with the Cassius "become him" resolution for oxygen. The recommendation trades a *catchable* reveal risk for the avoidance of an *un-catchable* moral-north corruption and a missed hard constraint. If Josh weights reveal-safety above constraint (2) and above the moral-north risk, Seating B (lock-respecting) is the ranked fallback — it is the second-best option in one review and first in the other, and no reviewer places it last.
---
On a ruling for the recommended A+Ch-65-seed configuration (adjust the CH_NN targets if Josh rules a different seating):
Worldstate wiring
FORK_CH75_D — anchor a new EVIL-only beat CH75_B04a ("the unarmed line"). magnitude SUBSTANTIAL · kind path_dispatch (band-gated) [moral_resolution if lock respected / combat_resolution if overridden] · care_band MAXIMUM · convergence_node null · progression_guarantee: no fail-wall, forward passage to Ch 76 never gates.ws.path_bucket (fires only == EVIL); ws.path_lock_in_state: LOCKED; optionally ws.familiar_bond_state for aftermath texture (how alone the player already is when she appears).ws.ntala_inversion_state: { STANDING_WITNESS | SILENCED } (new durable axis, F1 nemesis_memory-class, read at the Epilogue). Holds ws.journal_status: NEVER_COMPLETED + ws.journal_voice_texture: SPARSE_COLD band-fixed (spare/kill does NOT move them).ws.path_bucket (locked Ch 65), ws.companion_roster_state (owned by the terminal-stand / coalition forks), ws.father_encounter_variant (Ch-65 lock), ws.journal_status (band-fixed). Not wired to any construct — golem-exclusion (BE_0075.P3 has no director, CH_75.md:141).SC_CH65_threshold_crossing (CH65_B10), writing an inversion_begins state (and, pending the D-5 ruling, the record's crossing setup).CH_NN edits (all [JOSH]-gated on seating)
CH_65.md — replace the bare #146 "Ntala departs" EVIL cell with the staged inversion-seed beat; resolve the #252 "held in the world / audit-pending D-23" seam against the ruling.CH_75.md — add CH75_B04a + the FORK_CH75_D 2-row EVIL sub-fork (spared / felled); confirm FORK_CH75_A remains 6 rows untouched.CH_EPILOGUE (aftermath texture only) — journal-custody colouring per ws.ntala_inversion_state, with no reader identity assigned and journal_status held fixed (D-8).Registry rows
ws.ntala_inversion_state ∈ { STANDING_WITNESS | SILENCED }; the FORK_CH75_D row in the fork registry; optional parallel ws.relationship_history value.T3_Core_Characters §3.4 — only if Josh overrides the lock (D-2/D-6): author the social-derived, no-healing combat capability as an explicit, ruled invention; and regardless of the lock, resolve the stale line 203 "possible departure at Ch 23" language and the D-23 audit-pending baseline (D-7).Gates
CH_75.md:235 guardrail).Cardinal items requiring Josh's ruling before authoring: seating chapters (D-1) · the §3.4 override vs respect and any combat-ability grant (D-2/D-6) · the realm-transit crossing mechanism (D-5) · the T3 baseline / D-23 ratification (D-7). All marked [JOSH]; none asserted as decided.