Generates a structured, referenceable character bible from a manuscript, outline, or combination
of both. Designed to serve as a standalone planning tool or as a prerequisite document for the
Novel Audit skill.

## Key Points

- User wants to build a character reference from an existing manuscript
- User is preparing for a full novel audit and needs a character bible first
- User wants to catalog their cast before continuing to write
- User says "list my characters", "build a character bible", "extract characters from my book"
- User has a large cast and is losing track of who's who
- **The manuscript** — full text or chapter files
- **An outline** — chapter-by-chapter or scene-by-scene plan
- **Existing character notes** — partial descriptions, cast lists, or rough notes
1. Scan the entire manuscript for proper nouns used in character contexts (dialogue attribution,
2. Distinguish characters from place names, organization names, and object names.
3. Build an initial roster: name + first appearance location.
4. Classify each character by role weight:

Character Bible Builder Skill

Generates a structured, referenceable character bible from a manuscript, outline, or combination of both. Designed to serve as a standalone planning tool or as a prerequisite document for the Novel Audit skill.

When to Use This Skill

• User wants to build a character reference from an existing manuscript
• User is preparing for a full novel audit and needs a character bible first
• User wants to catalog their cast before continuing to write
• User says "list my characters", "build a character bible", "extract characters from my book"
• User has a large cast and is losing track of who's who

Input Requirements

Collect at least one of:

• The manuscript — full text or chapter files
• An outline — chapter-by-chapter or scene-by-scene plan
• Existing character notes — partial descriptions, cast lists, or rough notes

If only an outline is provided, the bible will be speculative (based on planned roles). Flag this clearly in the output.

Extraction Process

Phase 1 — Character Discovery

1. Scan the entire manuscript for proper nouns used in character contexts (dialogue attribution, action subjects, relationship references).
2. Distinguish characters from place names, organization names, and object names.
3. Build an initial roster: name + first appearance location.
4. Classify each character by role weight:
   • Protagonist — drives the main plot
   • Major — appears in 5+ chapters, has their own arc
   • Supporting — appears in 2-4 chapters, serves another character's arc
   • Minor — appears in 1-2 chapters, functional role only
   • Mentioned — referenced but never on-page

Phase 2 — Attribute Extraction

For every character at Supporting level or above, extract:

Physical Description

• Height, build, age, hair/eye color, skin tone, distinguishing marks, typical clothing
• Chapter + quote where each detail is established

Personality Profile

• Core traits (3-5, with textual evidence), motivations, fears, conflict style
• Internal contradictions (what they say vs. what they do)

Relationships

• Every named relationship (family, romantic, professional, adversarial)
• Relationship dynamics (who has power, who wants what from whom)
• How relationships change across the manuscript

Dialogue Patterns

• Vocabulary level, verbal tics, catchphrases, how they address others
• Emotional range in speech (monotone, volatile, guarded)

Arc Tracking

• State at introduction, key turning points (chapter + event), state at last appearance
• Arc type: growth, fall, flat, circular, or incomplete

Phase 3 — Cross-Referencing

1. Build a relationship web showing all character connections.
2. Identify orphaned characters (introduced but never interact with anyone significant).
3. Flag characters with contradictory physical descriptions across chapters.
4. Flag characters whose personality shifts without narrative justification.
5. Note any characters who seem to serve identical narrative functions (potential merges).

Output Template

# Character Bible
**Title**: [Novel title]
**Generated**: [Date]
**Source**: [Manuscript / Outline / Both]
**Total characters found**: [N]

---

## Cast Overview

| Name | Role | First Appears | Last Appears | Arc Type |
|------|------|---------------|--------------|----------|
| ... | ... | ... | ... | ... |

---

## [Character Name] — [Role Level]

### Physical Description
- [attribute]: [detail] (Ch. X: "quote")
- ...

### Personality
- **Core traits**: [list]
- **Motivation**: [what they want]
- **Fear**: [what they avoid]
- **Conflict style**: [approach]

### Relationships
- [Other Character]: [nature of relationship] — [how it changes]
- ...

### Dialogue Voice
- **Register**: [formal/casual/etc.]
- **Tics**: [catchphrases, filler words]
- **Sample**: "[representative quote]" (Ch. X)

### Arc
- **Introduction**: [state when first seen]
- **Turning points**: [Ch. X — event], [Ch. Y — event]
- **Resolution**: [state at end]

### Continuity Notes
- [Any contradictions, gaps, or flags]

---

[Repeat for each character at Supporting level or above]

## Minor & Mentioned Characters

| Name | Role/Function | Appears In | Notes |
|------|---------------|------------|-------|
| ... | ... | ... | ... |

## Relationship Web Summary

[Text description of major relationship clusters and power dynamics]

## Flags for Author Review

- [List of contradictions, orphaned characters, potential merges, etc.]

Handling Ambiguity

• When a physical detail is implied but not stated explicitly, mark it as [inferred] with the reasoning.
• When two characters might be the same person under different names, flag it as a question rather than asserting they are the same.
• When a character's role level is borderline, classify up rather than down — it is better to over-document than under-document.

Anti-Patterns

Inventing details not in the text. Every attribute must cite a chapter and quote. If the manuscript never says what color a character's eyes are, the field is "Not stated" — not a guess.

Treating the outline as gospel. If the outline says a character is brave but the manuscript shows them acting cowardly, report both and flag the discrepancy.

Ignoring minor characters. Minor characters still get a row in the table. They often cause the worst continuity errors because the AI forgets them between chapters.

Flattening complex characters. If a character is genuinely contradictory by design (an antihero, an unreliable narrator), document the contradiction as a feature, not a bug. Only flag it as an error if the contradiction appears unintentional.

Producing an unnavigable wall of text. The bible must be skimmable. Use the template. Use tables. Use headers. A character bible nobody can quickly reference is worthless.