book-studio
This is a starting point for handing book writing to AI and finishing it one chapter at a time. On the assumption that you work in Claude Code or Codex, the 4-Tier pipeline, the per-chapter verification loop, and a folder structure for running several books at once are laid out ahead of time. That keeps AI from losing the tone and continuity that long-form writing tends to drift on.
This page is just a fork in the road. Use the table below to figure out where you should start, then jump straight there.
Where to start
Section titled “Where to start”| If you are | Go to |
|---|---|
| Not a full-time developer, new to the terminal and git, new to book-writing automation | Beginner manual |
| Comfortable with code, but new to the 4-Tier flow, book_type routing, and the continuity index | The beginner manual from prerequisites through first-run, then the expert manual |
| Already worked on a book with an AI writing tool | Expert manual, straight to architecture |
If you can’t decide which row you’re in, pick the first one. The beginner manual moves one action per step, so you can skim past anything already familiar.
How the two manuals differ
Section titled “How the two manuals differ”The expert manual reads at reference density. It assumes the vocabulary of 4-Tier, the book_type skill graph, the per-chapter AI-smell gate, the continuity index, and multi-book is familiar, and it leads with the reasoning behind decisions. Six sections: intro, architecture, genres, verification, manuscript, updates.
The beginner manual is step by step. It assumes you’re comfortable with the idea of AI writing for you, but that the terminal, git, and markdown might be new. The text tells you what the screen should look like when things are working. It runs as an index plus twelve steps.
The two are the same base told at two paces. The tone differs, but the commands and workflow are identical.
Common to both
Section titled “Common to both”The first run barely touches external services. Call 프로젝트 시작 and /start checks only git, captures the book identity (type, logline, persona, tone), and hands off to Tier 1 of book-orchestrator. You need no API key or payment account at that point. External search or illustration reaches outside only when you explicitly request it.
Writing flows in 4 Tiers. Direction sets the direction, Planning lays out the outline and the world, Writing produces one chapter at a time, and Quality verifies. Because each Tier hands off through _workspace/{book_slug}/ files rather than chat history, a chapter survives a session break. You can resume days later with one sentence.
Every chapter has to clear the AI-smell gate to be marked finished. Character appearance, world rules, and the timeline are extracted into consistency_index.json and cross-checked at each chapter. It’s the spot that keeps a character with blue eyes in Chapter 3 from turning brown in Chapter 12.
The AI entry point is one line, /start, in Claude Code, and the repository root AGENTS.md in Codex. Both paths point at the same canonical document, AI_AUTOMATION.md.
The pipeline only produces per-chapter outputs. Combining the chapters that passed into a single manuscript, along with submission and ebook upload, is done by you. Nothing gets uploaded to an external platform automatically. That’s security baseline S8.
The license is All Rights Reserved with a commercial license model. The payment flow and a lawyer-reviewed COMMERCIAL-LICENSE.md are still in progress, so both need to be finished before real sales. The manual and landing are public, so the page you’re reading right now is open as is.
Once you’ve picked your row, jump straight to the expert manual or the beginner manual.