Awesome Novel Studio

April 2, 2026 · View on GitHub

Awesome Novel Studio

AI Web Novel Creation Harness — A Claude Code Plugin

English | 한국어

An end-to-end web novel production system built on Claude Code. Combines 18 specialist agents and 10 skills to automate the full pipeline: propose → design → create → polish → rewrite.

Production-Proven — A web novel written with this workflow secured a publishing deal. Daily views 2,500+ / Likes 1,000+ / Subscribers 300+

Installation

Via Marketplace

Add the marketplace

/plugin marketplace add MJbae/awesome-novel-studio

Install the plugin

/plugin install novel-studio@awesome-ai-studio

Restart session to activate

/exit

Quick Start

# 1. Generate 3 novel proposals, pick one
/propose

# 2. Big design (bootstrap + character sheet + plot-hook guide)
/design-big

# 3. Small design (25-episode detailed design)
/design-small

# 4. Write episodes
/create

# 5. Polish
/polish

Pipeline

Propose ────── Design ─────────────── Create ──── Polish ──── Publish
/propose        /design-big                /create   /polish
                /design-small
                     ↕ On design changes
                  /rewrite → /polish

The Three Walls of Long-Form AI Fiction

Writing a single 5,000-character episode with AI is easy. Scaling that to 300 episodes and 1.5 million characters is where things break.

The Wall	Symptom	Awesome Novel Studio's Solution
Character depth collapse	Distinct personalities, speech patterns, and motivations blur into sameness as episodes accumulate.	Character sheet + voice table — Define each character's tone, sentence endings, and non-verbal palette at design time. Automatically verified every episode via the `VOICE`, `TITLE`, and `ALIVE` polish axes.
Story coherence breakdown	Cause-and-effect chains and foreshadowing unravel across a sprawling world.	Plot-hook guide + continuity-bridge — Decompose the full narrative arc into 25-episode beat structures. Before each episode, `continuity-bridge` collects timeline, foreshadowing, and character state from the previous 2 episodes and feeds it to the creation agent.
Number and setting inconsistency	Currency values, historical dates, character ages — the hard facts that anchor the story start contradicting each other.	guard_rails + LOGIC axis — Define absolute rules in `novel-config.md`. The `LOGIC` polish axis cross-checks numbers and timeline against the previous 2 episodes.

Commands

Command	Description	Notes
`/propose`	Generate 3 proposals from genre + Korean platform + concept	Starting point for a new novel
`/design-big`	Full novel design (bootstrap, characters, plot)	Includes auto-research
`/design-small`	25-episode detailed design	Requires big design first
`/design`	Design router (big + small combined)	Use when scope is unclear
`/bootstrap`	Generate bootstrap document only	World-building and concept only
`/character`	Generate character sheet only	Character design only
`/plot-hook`	Generate plot-hook guide only	Narrative structure only
`/create`	Write episodes sequentially	Based on design documents
`/polish` (= `/lint`)	16-axis polish (6 agents in parallel)	Auto-sequential progression
`/rewrite` (= `/revise`)	Rewrite episodes after design changes	Auto-calculates impact scope

Scope Targeting

All create/polish/rewrite commands accept episode range arguments:

/create EP051          # Start from EP051
/create EP001-EP010    # EP001 through EP010
/polish start          # From EP001
/rewrite project-name EP001-EP010

Workflow

/propose                Generate 3 proposals, select 1
       ↓
/design-big             Bootstrap + character sheet + plot-hook guide
       ↓                → novel-config.md auto-generated
/design-small           25-episode detailed design (optional, recommended)
       ↓
/create                 Episode creation (auto-sequential)
       ↓
/polish                 16-axis polish (auto-sequential)
       ↓
[On design changes] /rewrite → /polish   Rewrite then re-polish

Directory Structure

Running the pipeline generates this project structure:

{project-name}/
├── novel-config.md              # Project config (central file)
├── design/                      # Design documents
│   ├── {name}_bootstrap.md      # World-building, concept, platform strategy
│   ├── {name}_character.md      # Character profiles, relationships, dialogue DNA
│   └── {name}_plot-hook.md      # 3-act structure, 25-ep beats, hook strategy
├── episode/                     # Episode manuscripts
│   ├── ep001.md
│   └── ...
├── revision/                    # Working files
│   ├── fix_plan.md              # Polish progress tracker
│   └── learnings.md             # Patterns discovered during polish
└── _workspace/                  # Temporary workspace
    └── 00_research/             # Auto-research results

novel-config.md

The central configuration file for the entire pipeline. Auto-generated by /design-big, then reviewed and edited by the user.

project:
  name: "My Novel"
  target_platform: "문피아"
  target_genre: "regression+specialist"
  episode_dir: "episode/"
  work_dir: "revision/"
  design_dir: "design/"

design_documents:
  bootstrap: "design/my-novel_bootstrap.md"
  character_core: "design/my-novel_character.md"
  plot_guide: "design/my-novel_plot-hook.md"

ep_range_table:
  - range: "EP001-EP025"
    label: "Act 1: Origin"
    plot_guide: "design/my-novel_plot-hook.md"

guard_rails:
  - "The protagonist's regression ability only allows recalling past information"

custom_axes:
  EXPERTISE: "Domain knowledge should be hinted at through dialogue, never explained"

ep_range_table: Episode ranges auto-extracted from the plot guide
guard_rails: Absolute rules enforced at every creation and polish step
custom_axes: Project-specific additional polish criteria

16-Axis Polish System

Axis	Name	Description
1	BANNED	Banned expressions (time-skip cliches, meta-commentary)
2	VOICE	Character dialogue consistency (checked against voice table)
3	TITLE	Forms of address (speaker, listener, context)
4	SILENCE	Silence pattern overuse (max 4 per episode)
5	TRANS	Translationese, AI tone, semantic literal translation detection
6	SCENE	Scene structure (beats, conflict, resolution)
7	LOGIC	Narrative logic, timeline, numerical consistency
8	SUMMARY	Scene-level justification of existence
9	UNIFORM	Cross-episode consistency
10	HOOK	Hook intensity (opening, midpoint, cliffhanger)
11	OPENING	Opening hook within 200 characters
12	MOBILE	Mobile readability (paragraph length, dialogue ratio)
A1	ALIVE	Echo dialogue resolution
A2	ALIVE	Silence-to-nonverbal replacement
A3	ALIVE	Character vitality at tension points
A4	ALIVE	Emotional distance management

Agent Architecture

Design Agents (5)

concept-builder · character-architect · plot-hook-engineer · proposal-generator · domain-researcher

Creation Agents (4)

episode-architect · episode-creator · continuity-bridge · quality-verifier

Polish Agents (6)

rule-checker · story-analyst · platform-optimizer · alive-enhancer · revision-executor · revision-reviewer

Rewrite Agents (4)

revision-analyst · character-sculptor · episode-rewriter · quality-verifier

quality-verifier is shared between Creation and Rewrite phases (18 unique agents total).

Special thanks to Minho Hwang (revfactory) — his Harness plugin made it easy to bootstrap the initial harness architecture. The insights he shares on LinkedIn and other channels have been a constant source of inspiration.

License

Apache 2.0