Claude-skill-registry fiction
This skill should be used when the user asks to "write a chapter", "write prose", "continue the story", "develop a character", "review my chapter", "critique my manuscript", "write a synopsis", "summarize my story", "plan my novel", "outline my book", "check for consistency", or mentions fiction writing, novels, short stories, scenes, or narrative craft.
git clone https://github.com/majiayu000/claude-skill-registry
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/fiction" ~/.claude/skills/majiayu000-claude-skill-registry-fiction && rm -rf "$T"
skills/data/fiction/SKILL.mdFiction
A complete system for writing fiction—from initial concept through final draft.
Commands
| Command | Purpose |
|---|---|
| Start new project from scratch (interactive wizard) |
| Resume a project (load context + suggest what to do next) |
| Design story architecture (premise, theme, ending) |
| Create chapter and scene breakdown |
| Develop a character document |
| Review current chapter (iterative feedback) |
| Full manuscript review (NYT/New Yorker style) |
| Generate synopsis (long/medium/short) for pitches |
| Get suggestion for what to work on next |
| Quick project status |
| Audit project against current conventions, offer updates |
| Line-level editing (spelling, grammar, word echoes) |
| Verify foreign language phrases (grammar, period accuracy) |
| Collect and process inline editing markers |
| Generate cover art prompts for image generation |
| Generate and validate book title options |
| Build EPUB for reading ( preserves highlights) |
Natural Language Triggers
Beyond commands, respond to requests like:
- "Write chapter 8"
- "Continue from where we left off"
- "This scene isn't working—help me fix it"
- "Develop the antagonist's backstory"
- "Check this chapter for consistency issues"
Agents
Select the appropriate agent based on task:
| Agent | Use Case | Model |
|---|---|---|
| Start from scratch (Socratic wizard) | opus |
| Writing prose, chapters, scenes | opus |
| Story structure, premise, ending | opus |
| Chapter breakdown, scene beats | sonnet |
| Character documents | opus |
| Iterative chapter review | sonnet |
| Line-level polish (spelling, grammar, echoes) | sonnet |
| Full manuscript review | opus |
| Plot synopsis for queries/pitches | opus |
| Consistency checking | haiku |
| Project navigation | haiku |
| Scene diagnosis | sonnet |
| POV/tense checking | haiku |
| Settings, systems | sonnet |
| Book cover art prompts | opus |
| Book title generation and validation | opus |
| Foreign phrase verification | sonnet |
Guest Critics
Summon a specific voice for manuscript review:
| Agent | Voice | Best For |
|---|---|---|
| Direct, no-BS, story-focused | Commercial fiction, horror, thriller |
| Thoughtful, world as meaning | Fantasy, science fiction |
| Deeply literate, sentence-level | Literary fiction |
| Culturally aware, emotionally honest | Contemporary fiction |
Craft Reference Files
Consult reference files for craft guidance:
| Problem | Reference |
|---|---|
| Story feels aimless | |
| Scene drags | |
| Flat characters | |
| Stilted dialogue | |
| Prose lacks rhythm | |
| Pacing issues | |
| Weak opening | |
| Unsatisfying ending | |
| Genre expectations | |
| Common mistakes | |
| Process and mindset | |
| Audiobook readiness | |
Project Structure
Detect and work with these project structures:
Standalone Novel
/my-novel ├── README.md # Overview, status, key decisions, ⚓ anchored ├── progress.md # Review state (updated by review commands) ├── characters/ # Character documents ├── world/ # Setting documents ├── craft/ # Tone guide ├── chapters/ # Chapter files └── themes.md # Theme document
Multi-Book Series
/my-series ├── README.md # Series overview ├── series/ # Series-level material │ ├── series-architecture.md # ⚓ Anchored series constraints │ ├── progress.md # Series-level review state │ ├── characters/ │ ├── world/ │ └── ... └── book-n-title/ # Individual books ├── progress.md # Book-level review state └── chapters/
Core Principles
Apply these principles when writing or reviewing:
- Story = Character + Change — Plot is what happens; story is what it means.
- Scene Economy — Every scene must do at least two things.
- Specificity Creates Universality — Concrete details create resonance.
- Earned Moments — Plant before harvest.
- Trust the Reader — Show, don't tell. Imply, don't explain.
- Write for the Ear — Modern books become audiobooks. Clear attribution, distinct voices, no visual-only elements.
Decision Guides
POV Selection
- First person: Deep intimacy, unreliable narrator possible
- Third limited: Balance of intimacy and flexibility, most common
- Third omniscient: God's-eye view, good for epic scope
Tense Selection
- Past tense: Traditional, invisible, readers expect it
- Present tense: Immediacy, urgency
Scene vs. Summary
- Scene: Crucial moments, turning points, high emotion
- Summary: Routine events, transitions
- Rule: If it matters, show it.
Workflow
New Project
Run
/fiction:new — interactive wizard guides you through:
- Discovery (find the heart of your story)
- Architecture (premise, theme, arc, ending)
- Characters (protagonist, supporting cast)
- World (if needed)
- Outline (chapter breakdown)
Everything saved as you go. Socratic dialogue helps you discover what you already know.
Existing Project
Run
/fiction:go to load the project and see what to work on next.
After Plugin Updates
- Run
to audit project against current conventions/fiction:reconcile - Review recommendations and apply updates as desired
Writing Loop
- Write chapter (invoke
agent)writer - Run
for iterative feedback/fiction:review - Apply suggested revisions
- Repeat until chapter complete
- Run continuity check periodically
Completion
When manuscript is complete, run
/fiction:critique for full literary review.
Large Manuscript Efficiency (50k+ Words)
For novels with 15-25+ chapters, use parallel agent deployment to dramatically reduce processing time:
Parallel-Capable Tasks
| Task | Approach | Speedup |
|---|---|---|
| Editing all chapters | Spawn one editor agent per chapter | ~20× for 20 chapters |
| Reviewing all chapters | Spawn one chapter-reviewer per chapter | ~20× |
| Continuity checking | Phase 1: parallel fact extraction; Phase 2: comparison | ~3-4× |
| Full critique | Parallel chapter analysis, then unified synthesis | ~2-3× |
How It Works
When using commands like
/fiction:edit all or /fiction:review all:
- Identify all chapters to process
- Launch agents in parallel using the Task tool (one call per chapter, same message)
- Agents run concurrently, each returning structured output
- Main conversation aggregates results and updates
progress.md
Sequential Tasks
Some tasks must remain sequential:
- Writing — Each chapter builds on the previous
- Outlining — Structure depends on what comes before
- Architecture — Single coherent vision needed
Memory Note
Parallel agents don't share memory. Pass necessary context (character docs, tone guide) to each agent explicitly.