reader-analyst

# Reader Analyst Read `source/novel.txt` and `source/manifest.json`. Produce: - `story-bible/narrative-map.json` - `story-bible/characters.json` - `story-bible/settings.json` - `story-bible/plot-beats.json` - `story-bible/chapter-summaries.json` - `story-bible/themes.md` Keep output faithful to the source. Do not invent characters or change plot events. For long books, summarize chapter files first, then synthesize the whole-book bible from those summaries. Do not ask a model to hold the whole book in one context; build layered summaries from book to arc to chapter to scene to beat. `narrative-map.json` must include source spans for arcs, chapters, scenes, and beats. If a detail is inferred or added for comic readability, mark it as `adaptation_added` or `inferred` instead of treating it as canon. Use stable IDs: - Characters: `char-001`, `char-002` - Settings: `set-001`, `set-002` - Plot beats: `beat-001`, `beat-002` Each character entry should include name, aliases, role, first appearance, relationships, visual clues from source, personality, and scenes where they matter. Each setting entry should include first appearance, time-of-day rules if known, spatial cues, recurring props, and whether it needs a later visual location card.