decisions/OPEN_DECISIONS_BRIEF.md
Produced by a deep-dive + adversarial-review Workflow (wf_936413bb; 9 items × research → adversarial → finalize → synthesis, 28 agents), each fact-checked against the actual canon (CVD, tier docs, RCM, registries, spines) AND real-world sources (WebSearch), then challenged before a recommendation was finalized. Full per-item packages (conflict / evidence / options / apply-cost / confidence) are in the workflow output; this is the decision layer. Nothing below is applied — these await your call.
Corrections the deep-dive found that hold under ANY option (objective, not your cardinal call — I can apply on your word): the D-10 flag's "d. 2014 / recently-deceased" is FALSE (the Dambana chief is living — met Sri Lanka's PM 2026-06-13); the D-11 site carries fabricated Aksum coordinates + a false Ark-of-the-Covenant weld; the Ch 10 [SRC: EXPANDED_COVERAGE_MAP # A1] cite is wrong (A1 is the Ch-5 Angkor leg); Ch 14 Arikhankharer never reigned ("king" → "royal prince"); the Ch 13 glyph ‡ (U+2021) should be ǂ (U+01C2) with the right click gloss; enrich.py's docstring still re-seeds "Murugan F2" into every build pack.
---
Living Ju\|'hoansi (Kalahari) foragers + an irreducibly Ju\|'hoan boss sit physically in the East African Rift, whose real present-day forager — the Hadza (~1,000 people, Lake Eyasi, ~90% land lost in 50 years) — appears nowhere in all 79 nodes.
RECOMMENDATION: (c), using (e)'s time-fluctuation vision as the reframe mechanism. Lead with Hadza; hold Sandawe optional. The Hadza are the Rift's real, pressured forager and a genuine §17.1 Care gap; the boss is Ju\|'hoan-specific by attestation so it can't be relabeled Hadza — (c) keeps it as an honest echo instead of pretending the transplant is gone. *Confidence: High. Two hard guardrails: Hadza is a language isolate (any click-sharing note is areal only, never San kinship — the "Khoisan" macro-family is rejected), and Hadza content must never read as "living fossils."*
"Uru Warige Wanniya" borrows the living Dambana chief's community self-designation onto an invented supernatural NPC. (The recognition-power repair — reframed as human perceptiveness under §17.9 — is already applied and stands.)
RECOMMENDATION: (a-prime) — drop the "Wanniya" styling, leave the specialist un-named, keep "of the Uru-Warige clan" as optional texture; (a) is the fallback if you want a name. The referent is living and active, which removes any pull toward homage; "Uru Warige" is a shared matrilineal clan, so dropping the personal self-designation severs the specific real-person link. *Confidence: High. Mandatory regardless: correct the false "d. 2014" fact in all four docs.*
Majapahit didn't exist until 1293 and reached Bali only in 1342–43 (~90–143 years off); every real Ch-4 anchor site (Blanjong 914, Tirta Empul 962, Gunung Kawi ~1080) is Warmadewa/Bali-Kingdom.
RECOMMENDATION: (c) — "1200 CE — Bali Kingdom (Warmadewa dynasty); Majapahit-absorption stratum reached across time-fluctuation windows," mirroring Ch 5's phrasing, with a Blanjong-anchored (914 CE) first-writing rider. 1200 CE Bali is documented Warmadewa fact, neighbors Ch 3/5/6 already use composite eras, and Majapahit does no plot work (the editing-hand theme rides the fictional NW_0002 apparatus) — so this improves fidelity at the lowest ripple. *Confidence: High. Only the label wording is open. Ch 5's Majapahit composite is NOT the same anomaly — no matching edit there.*
---
The anchor names a temple 7 km from Marib, Yemen, but pins it to fabricated Aksum coordinates (~650 km misplaced) and wrongly welds it to the Ark tradition in six ACTIVE registries. The built spine already sits at the hybrid framing.
RECOMMENDATION: (c) — ratify the hybrid the spine already embodies + propagate to the six registries. Yeha is Ethiopia's oldest standing structure (~700 BCE), a Sabaean-style Almaqah temple that fits the Aksum-region chapter where the Yemeni Marib temple cannot; naming Marib one layer out satisfies represent-with-care. *Confidence: High. Use the verified Yeha coords (~14.27°N/39.05°E), never a placeholder; the GE_EZ→Ezana-stele repoint is a real catch — don't naively swap that token to "Yeha."*
"Lebogang" is a Tswana/Sotho name — the wrong linguistic family for a Maa moran.
RECOMMENDATION: (b) — rename to Lemayian ("the blessed one"), an attested Maa given name, keeping the moran role + the code-switch field. Prefer a given name over an *ole*-patronymic; Sironka is borne by a known contemporary Maasai artist, so Lemayian is the safer default. *Confidence: High. One final real-person screen before commit. Independent of the geography choice.*
The NPC name leads with "Ama-" and reads as Amma (the creator deity his office serves) to any non-specialist.
RECOMMENDATION: (b) — reframe, as a precautionary confirm consistent with Senami/Gbetosi/Kaggen (NOT a finding that "Amadigne" is a true theophoric). §17.1 is governed by player-facing surface, not specialist linguistics — "Ama-" reads as Amma and the geminate distinction is invisible in play; only a non-combat teaching-NPC name moves (Amma/Nommo/Sirius content all stand). *Confidence: High. The sweep is 9 files (not 5); grep "Amadigne" → 0 live hits before done.*
Four §17.1 reframes applied per the precedent chain; Ch 14 overstates a real prince's rank as "king," and Ch 14/15 defer their final constructed names.
RECOMMENDATION: (a). Each reframe maps to a Josh-ruled precedent; the one genuine error is factual — Arikhankharer was the eldest son who predeceased his parents and never reigned, so "king" overstates rank. *Confidence: High. Non-blocking residuals routed to NAMING_METHOD: (1) Ch 14 "the Kandake" must read as one corrupted individual, never an indictment of the revered institution; (2) the constructed replacements aren't screened yet (the reframe already threw two real-king collisions — working names aren't final); (3) fix the Ch 13 glyph ‡→ǂ + its wrong click gloss.*
Should the Manu–Matsya flood myth and Indian numerals/zero home to Ch 16 (Indus/North) or Ch 7 (Tamil South)? Build + link-graph anchor already read Ch 16.
RECOMMENDATION: (a) — confirm both at Ch 16. The Matsya myth's earliest attestation is the North-Indian Shatapatha Brahmana and the numerals are classical North-Central (Aryabhata at Pataliputra, Brahmagupta at Bhinmal) — neither Harappan nor Tamil; moving them to Dravidian Ch 7 would introduce the exact appropriation-confusion Care guards against. *Confidence: High. If you ever want Manu shown living, its home is Ch 5/6, never Ch 7; the true-origin Gupta node is a separate new-chapter call.*
The controllable F2 peacock was named "Murugan" (deity name on a mortal entity — the Kxao-rule breach); the rename to Paravani landed in only one of the files carrying it.
RECOMMENDATION: (a) — confirm Paravani, five-file widened fix, enrich.py the priority edit. Paravani is the attested name of Murugan's peacock vahana, and its in-world origin as a subjugated asura remade into the mount clears the Kxao line decisively; the deity stays honored + non-target. All targets are _source living docs + one docstring — no fidelity re-baseline. *Confidence: High. For awareness only (no action): the Bond-100 ultimate names the ability "Kartikeya … activation" — a reverent iconographic register, not the entity's name, so defensible as-is.*
Do the two edges seed the ~Ch 38 pattern payoff without letting the player connect the recognitions into one agent/family before its gate? (Ch 18's is the D-08 L1 hardening you ratified 2026-07-08.)
RECOMMENDATION: (a) — confirm both, no file edit. The forbidden move is a narrator claim that the recognitions are one connected agent; Ch 18 asserts only method-recurrence (agent left empty), and Ch 19 lands "cannot yet name or connect" in the negative at the exact enumeration moment. Both sit ~20 chapters under the bar and satisfy §17.9 trivially. *Confidence: High. For your eye only: Ch 18 L30 ("a hand that wanted to be the only one still holding the pen") is the second-highest-implication phrase — still under the line, but the one to judge if you want to gauge proximity. The out-of-context-clip/marketing risk on the scripture-editing premise remains a separate QA/comms call.*