/crucible-review

Trigger a manual review of your chapters using specialized review agents.

Usage

• /crucible-suite:crucible-review latest - Review most recent 2 chapters
• /crucible-suite:crucible-review 5-6 - Review chapters 5 and 6
• /crucible-suite:crucible-review 1-4 - Review chapters 1 through 4

Execution Instructions

IMPORTANT: When this command is invoked, you MUST:

0. ALWAYS use the AskUserQuestion tool for presenting options to the user (NOT plain text A/B/C options)

1. Detect project root - Find the Crucible project directory:

   • Use Glob to search for .crucible/ directory: Glob(pattern="**/.crucible")
   • The project root is the PARENT of the .crucible/ directory found
   • Store this as PROJECT_ROOT (absolute path) for all subsequent steps
   • If no .crucible/ found, inform user this must be run from within a Crucible project

2. Determine chapter range from the argument:

   • latest → Read {PROJECT_ROOT}/.crucible/state/draft-state.json to find the last 2 completed chapters
   • X-Y → Use chapters X through Y
   • Single number → Review that chapter only

3. Resolve all file paths - Build absolute paths for agents:

   • Chapter files: {PROJECT_ROOT}/draft/chapters/ch{NN}.md
   • Story bible: {PROJECT_ROOT}/story-bible.json
   • Style profile: {PROJECT_ROOT}/style-profile.json
   • Master outline: {PROJECT_ROOT}/outline/master-outline.md
   • Scene breakdown: {PROJECT_ROOT}/outline/scene-breakdown.md
   • World forge: {PROJECT_ROOT}/planning/world-forge.md

4. Launch ALL 5 review agents in parallel using the Task tool with ABSOLUTE PATHS:

Task(subagent_type="crucible-suite:voice-checker", prompt="Review chapters X-Y for voice consistency.

PROJECT ROOT: {PROJECT_ROOT}

CHAPTERS TO REVIEW:
- {PROJECT_ROOT}/draft/chapters/ch05.md
- {PROJECT_ROOT}/draft/chapters/ch06.md

REFERENCE FILES:
- Style profile: {PROJECT_ROOT}/style-profile.json

Analyze for voice consistency against the style profile. Return a structured report following your output format.")

Task(subagent_type="crucible-suite:continuity-checker", prompt="Review chapters X-Y for continuity errors.

PROJECT ROOT: {PROJECT_ROOT}

CHAPTERS TO REVIEW:
- {PROJECT_ROOT}/draft/chapters/ch05.md
- {PROJECT_ROOT}/draft/chapters/ch06.md

REFERENCE FILES:
- Story bible: {PROJECT_ROOT}/story-bible.json
- Planning docs: {PROJECT_ROOT}/planning/

Check for continuity errors against the story bible. Return a structured report following your output format.")

Task(subagent_type="crucible-suite:outline-checker", prompt="Review chapters X-Y for outline adherence.

PROJECT ROOT: {PROJECT_ROOT}

CHAPTERS TO REVIEW:
- {PROJECT_ROOT}/draft/chapters/ch05.md
- {PROJECT_ROOT}/draft/chapters/ch06.md

REFERENCE FILES:
- Master outline: {PROJECT_ROOT}/outline/master-outline.md
- Scene breakdown: {PROJECT_ROOT}/outline/scene-breakdown.md
- Forge points: {PROJECT_ROOT}/planning/forge-points/

Verify adherence to the outline. Return a structured report following your output format.")

Task(subagent_type="crucible-suite:timeline-checker", prompt="Review chapters X-Y for timeline consistency.

PROJECT ROOT: {PROJECT_ROOT}

CHAPTERS TO REVIEW:
- {PROJECT_ROOT}/draft/chapters/ch05.md
- {PROJECT_ROOT}/draft/chapters/ch06.md

REFERENCE FILES:
- Story bible (timeline data): {PROJECT_ROOT}/story-bible.json
- World forge (travel times): {PROJECT_ROOT}/planning/world-forge.md

Analyze chronological consistency and temporal logic. Return a structured report following your output format.")

Task(subagent_type="crucible-suite:prose-checker", prompt="Review chapters X-Y for prose quality.

PROJECT ROOT: {PROJECT_ROOT}

CHAPTERS TO REVIEW:
- {PROJECT_ROOT}/draft/chapters/ch05.md
- {PROJECT_ROOT}/draft/chapters/ch06.md

REFERENCE FILES:
- Style profile: {PROJECT_ROOT}/style-profile.json

Analyze prose craft including pacing, show vs tell, and dialogue. Return a structured report following your output format.")

NOTE: Replace {PROJECT_ROOT} with the actual absolute path discovered in step 1, and adjust chapter numbers/filenames based on step 2.

5. Wait for all agents to complete and collect their reports

6. Compile a consolidated report showing:

   • All critical issues (from any agent)
   • All warnings (grouped by type)
   • All suggestions
   • Overall scores from each agent

7. Update the draft state to record the review using the update script:

python "${CLAUDE_PLUGIN_ROOT}/skills/crucible-writer/scripts/update_story_bible.py" "{PROJECT_ROOT}" --mark-review-complete <last_chapter_reviewed>

This sets review_pending = false and last_review_at_chapter to the last chapter number reviewed.

What Each Agent Does

1. Voice Checker (voice-checker)

• Analyzes voice consistency against your style sample
• Identifies tone shifts
• Checks POV adherence
• Evaluates character voice distinction

2. Continuity Checker (continuity-checker)

• Verifies character details match story bible
• Checks physical descriptions
• Validates relationship states
• Flags potential plot holes

3. Outline Checker (outline-checker)

• Verifies all planned scenes are present
• Checks beat coverage
• Identifies deviations from outline
• Notes acceptable creative changes

4. Timeline Checker (timeline-checker)

• Verifies chronological consistency
• Checks time references
• Validates travel times
• Tracks character age/state progression

5. Prose Checker (prose-checker)

• Evaluates show vs tell balance
• Identifies pacing issues
• Assesses dialogue effectiveness
• Highlights strongest passages

Review Output

Present a consolidated report:

═══════════════════════════════════════
BI-CHAPTER REVIEW COMPLETE
Chapters: [X-Y]
═══════════════════════════════════════

CRITICAL ISSUES (Must Fix)
─────────────────────────────────────
[Collected from all agents]

WARNINGS (Should Fix)
─────────────────────────────────────
Voice: [issues]
Continuity: [issues]
Outline: [issues]
Timeline: [issues]
Prose: [issues]

SUGGESTIONS (Consider)
─────────────────────────────────────
[Collected from all agents]

OVERALL SCORES
─────────────────────────────────────
Voice Consistency: X/10
Continuity: X/10
Outline Fidelity: X/10
Timeline: X/10
Prose Quality: X/10
AVERAGE: X/10

NEXT STEPS
─────────────────────────────────────

Then use AskUserQuestion:

{
  "questions": [
    {
      "header": "Next Steps",
      "question": "How would you like to proceed after this review?",
      "options": [
        {"label": "Address issues", "description": "Fix critical issues before continuing"},
        {"label": "Continue writing", "description": "Note issues and continue (return later)"},
        {"label": "Show details", "description": "View detailed report from specific agent"}
      ],
      "multiSelect": false
    }
  ]
}

When to Use

• When you want feedback before the automatic bi-chapter trigger
• After making significant revisions
• When something feels "off" but you can't identify it
• Before sending chapters to beta readers
• To check specific problem areas

Prerequisites

Requires written chapters to review. Works best with:

• Style profile on file (style-profile.json at project root)
• Story bible populated (story-bible.json at project root)
• Chapter outline available (outline/master-outline.md at project root)