narrative-authenticitylisted

# Narrative Authenticity AI-generated fiction is detectable at two independent layers, and only one of them can be fixed after drafting: 1. **Prose tells** — word choice, rhythm, cliché. Editable. 2. **Structural tells** — how the story is *conceived*: explained themes, tidy single-track plots, linear time, protagonist-choice resolutions. Research (StoryScope, arXiv:2604.03136) shows these survive professional style editing at ~94% detection. They must be planned out at outline time, never patched at edit time. ## Governing files (read before acting) - `_config/narrative_authenticity.md` — the full directive: Layer 1 structural rules, Layer 2 prose rules, Layer 3 Claude fingerprint counters. **This is the source of truth.** - `_config/narrative_audit_rubric.md` — 30-question structural self-check. - `stages/02_planning/output/structure_plan.md` — the project's chosen dials and the obligatory-scene ledger (if it exists). - `stages/01_onboarding/output/bible/genre_bible.md` + `setup/genre_bibles/INDEX.md` — the trope stack and reader contract. **Tropes outrank dials**: never delete, weaken, or "subvert" a ledgered obligatory beat; the authenticity rules apply to everything around those beats. ## What to do, by task ### Planning an outline or beats (Stage 02) Apply Layer 1 and produce/update `structure_plan.md` with: subplot map (≥2, not all theme-tight), nonlinearity plan, resolution variety table, moral-ambivalence beats, intertextual anchors, escalation contour, loose-e