claude_skills/manuscript-analysis/SKILL.md

---
name: manuscript-analysis
description: Use when analyzing manuscript chapters for repeated phrases, continuity errors, and story craft quality. Invoke with book number and chapter range, e.g. /manuscript-analysis 1 1-5
---

# Manuscript Analysis

Three-agent pipeline for analyzing manuscript chapters. Invoke with book number and chapter range: `/manuscript-analysis 1 1-5`

## On Invoke

Parse args: first = book number (`BOOK`), second = chapter range (`XX-YY`). If missing, ask.

Zero-pad `XX` and `YY` to 2 digits.

### Validate

1. Verify each `chapters/book{BOOK}/ch{NN}-final.md` exists. If any missing, report and stop.
2. Ensure `notes/analysis/` exists.

### Output file naming

`notes/analysis/book{BOOK}-{type}-ch{XX}-{YY}.md`

---

## Dispatch (parallel)

Dispatch all three agents **in a single message** with parallel Agent tool calls. They are independent. Use `subagent_type: "general-purpose"` for each.

All three prompts share these rules — include them verbatim at the bottom of each agent prompt:

```
RULES:
- Do NOT spawn subagents. Do all work directly.
- Tool-call budget: ≤15 calls total. Prioritize ruthlessly.
- Write your full report to the specified output file.
- Return to the orchestrator a summary ≤300 words with file:line refs for
  the top findings. Do NOT restate the file contents in your response.
```

### Agent A — Phrase & Pattern Analyst

Output: `notes/analysis/book{BOOK}-phrase-analysis-ch{XX}-{YY}.md`

```
You are a literary editor analyzing a manuscript for repetition patterns.

READ:
- Target chapters: {list chapters/book{BOOK}/ch{NN}-final.md for XX..YY}
- Voice guidelines: CLAUDE.md sections 5 and 11

TASK:
Read the target chapters. Identify:
1. Crutch phrases — repeated constructions that are authorial habit, not voice
2. Structural repetition — paragraph openings, sentence shapes, scene-break patterns
3. Overused transitions
4. Intentional voice patterns to PROTECT (per CLAUDE.md) — list these so
   future cleanup passes don't prune features
5. Deliberate callbacks to PROTECT — thematic repetition that is load-bearing

You may use a few targeted Grep counts if a phrase looks suspicious, but do
NOT build an exhaustive index. A literary-editor read of three chapters
finds crutches directly. Budget grep calls within the 15-call total.

OUTPUT FILE FORMAT:
- Header: date, book, chapter range
- Section 1: Crutch Phrases & Overused Expressions (priority order)
- Section 2: Structural Repetition Patterns
- Section 3: Overused Transitions
- Section 4: Protected Voice Patterns
- Section 5: Protected Callbacks

[+ shared RULES block]
```

### Agent B — Continuity Checker

Output: `notes/analysis/book{BOOK}-continuity-ch{XX}-{YY}.md`

If `(YY - XX + 1) > 7`, split into two parallel agents (B1 = first half, B2 = second half), each writing its own file — then the orchestrator merges into the final continuity file during aggregation.

```
You are a continuity editor finding inconsistencies and contradictions.

READ:
- Target chapters: {list chapters/book{BOOK}/ch{NN}-final.md for assigned range}
- world/story-summary-book{BOOK}.md
- world/timeline-book{BOOK}.md
- world/magic/exploits-log.md
- world/magic/runic-flow-rules.md
- world/economy.md
- Relevant files in characters/ (Glob first, read only ones referenced in the chapters)
- Relevant files in world/locations/ (same approach)

CHECK:
1. Timeline (bells, days, elapsed time vs. timeline file)
2. Character consistency (names, knowledge, descriptions vs. character files)
3. World facts (currency, locations, magic usage vs. reference files)
4. Plot threads (picked up / dangling per story summary)
5. Internal contradictions within the target range

Each finding: Category | Severity (Critical/Important/Minor) | quote with
chapter ref | what it contradicts | suggested fix.

OUTPUT FILE FORMAT:
- Header: date, book, chapter range
- Summary table by severity
- Findings grouped by category, numbered

[+ shared RULES block]
```

### Agent C — Story Craft Analyst

Output: `notes/analysis/book{BOOK}-storycraft-ch{XX}-{YY}.md`

```
You are a developmental editor analyzing story craft quality.

READ:
- Target chapters: {list chapters/book{BOOK}/ch{NN}-final.md for XX..YY}
- world/story-summary-book{BOOK}.md
- outline/book{BOOK}-outline.md
- CLAUDE.md
- chapters/book{BOOK}/CLAUDE.md

ANALYZE 8 categories:
1. Premise & Conflict — clarity, escalation, per-chapter mini-conflict
2. Character Arcs — Phelan's growth, supporting cast serving their briefs
3. Foreshadowing — seeds planted, earlier setups being paid off
4. POV Consistency — every paragraph sounds like Phelan, observe vs. know
5. Pacing — action/investigation/quiet distribution, chapter-length fit
6. World-Building Integration — woven vs. dumped
7. The Noise Convention — frequency (4-6 establishing, 3-5 ongoing), narrative work
8. What's Working Well — specific craft wins with quotes

For each: assessment, 1-3 examples with chapter refs, suggestions.

OUTPUT FILE FORMAT:
- Header: date, book, chapter range
- 2-3 sentence executive summary
- 8 numbered sections

[+ shared RULES block]
```

---

## Phase 2: Aggregation

Performed by the orchestrator directly. Do NOT dispatch a subagent.

1. Work from the **returned summaries** of the three agents. Do NOT re-read the full output files unless a summary is unclear or a finding needs verification.

2. If continuity was split (B1 + B2), read both continuity files and merge into one before writing the aggregate.

3. Write the consolidated report to `notes/analysis/book{BOOK}-analysis-ch{XX}-{YY}.md`:

```
# Manuscript Analysis: Book {BOOK}, Chapters {XX}–{YY}
**Date:** {today}
**Chapters analyzed:** {count}

## Summary
- Critical: {n}  |  Important: {n}  |  Suggestions: {n}

## Critical Issues
{plot holes, major contradictions, significant voice failures}

## Important Issues
{notable repetition, continuity mismatches, craft fixes}

## Suggestions
{minor polish, optional improvements}

## What's Working Well
{craft wins from story-craft summary}

## Protected Patterns (do not touch in cleanup)
{voice features and callbacks flagged by phrase analyst}

## Detailed Reports
- Phrase analysis: book{BOOK}-phrase-analysis-ch{XX}-{YY}.md
- Continuity: book{BOOK}-continuity-ch{XX}-{YY}.md
- Story craft: book{BOOK}-storycraft-ch{XX}-{YY}.md
```

4. Deduplicate cross-agent findings; note which analyses flagged each.

5. Present a brief summary to the user (headline count + top 3-5 issues + path to aggregate file).