How to use the story-setup agent skill for web novels

What the story-setup skill makes the agent do

story-setup deploys the web-fiction writing toolkit into your book project folder. It first works out which AI tool you are using, then follows the deployment steps for that tool. For tools with built-in support (Claude Code, OpenCode, Codex, Google Antigravity, ZCode, OpenClaw and Reasonix) it uses their specific hooks, agents and config. For Web AI or a generic agent, it uses a plain file mode built on skills and an AGENTS.md file.

The skill keeps your existing settings and manuscripts. Managed files are updated according to a manifest, and files shared with other tools are merged inside managed blocks. It does not overwrite your manuscripts, settings, outlines, tracking files or the .active-book file. After installing, it reports what it did and what to do next.

• Detects the host tool from your environment or project folder markers
• Copies toolkit skills, helper agents, hooks and rule files into the project
• Merges shared instruction files using managed blocks instead of replacing them
• Writes a .story-deployed sentinel file that records the installed state
• Runs a verification step and reports the deployment details

Install the story-setup skill

Option 1: download the ZIP from this page and unzip it into ~/.claude/skills/ for all your projects, or into a project's .claude/skills/ folder for one project only.

Option 2: in the Claude apps, open Settings, then Capabilities, then Skills, and upload the ZIP there.

If you installed the full toolkit from the command line and need to reinstall it, the source's own command is shown below. Use the marketplace or Plugin Management panel instead if that is how you installed it.

Trigger the skill with a request

You can run the skill with a slash or dollar command, or just ask in plain language. The source lists these triggers: /story-setup, $story-setup, and the Chinese phrases 准备写书 (prepare to write a book), 帮我搭一下环境 (help me set up the environment), 配置写作项目 (configure the writing project) and 检查写作环境 (check the writing environment).

To only check an existing installation without changing anything, ask it to check the deployment or diagnose the environment. The skill then runs its check-only path and reports the results without starting a deployment.

• /story-setup or $story-setup to start directly
• 准备写书 when you are starting a new book
• 检查写作环境 to check an existing setup without changing files

Example: setting up a new long-form project

Open your agent in the folder where your book will live. That folder becomes the project root, which is normally the current working directory. Then ask the agent to prepare the writing project.

If the agent cannot tell which tool you are using, it asks you in plain language. If the folder already has markers for another tool, it asks whether to update that tool too or only the one you use now. Answer those questions and let it finish. It will list the deployment details, including where files were placed.

• Project root: the folder you run the agent in, unless it is your home folder, a system folder or the skill's own folder
• Existing manuscripts, settings and tracking files are left alone

What the skill needs and what it changes

You need an AI agent that supports SKILL.md skills. The source lists Claude Code, Codex, OpenCode, Antigravity, ZCode, OpenClaw, Reasonix or a generic agent. Node.js is needed for the Antigravity agent generator script, and Python is needed for merge and copy-safety helper scripts. The skill runs scripts and does not use the network.

Because it merges configuration, you should understand the changes before you run it. It modifies files such as hooks and instruction files. If the reference bundle is incomplete or the tool is not supported, it stops and reports the missing paths instead of guessing.

• Runs scripts: yes (Python and, for Antigravity, Node.js)
• Uses network: no
• Stops without writing any deployment files if the bundle is incomplete
• Stops rather than downgrading if the project was deployed by a newer toolkit version

Safety notes before you run it

The installer copies files and merges settings into your project. Before running it on a project with custom hooks, agents or instruction files, back up the folder or put it under version control so you can compare the changes.

The skill is designed to be safe to re-run. Copy and merge steps repeat with the same result, so if an install is interrupted by a tool, permission or network problem, rerunning from the start is the recommended fix. It checks each copy source and target before copying to avoid copying a directory into itself.

• Keep a backup or version control before the first run
• Rerunning after an interruption is safe according to the skill's design
• Files it does not manage, such as your manuscripts and outlines, are not overwritten

Tips for getting the most out of it

Run the setup once per book project rather than in your home folder. If you maintain several book projects, run the same setup in each so they stay consistent after updates.

If you use more than one tool on the same folder, say so when asked, so the skill records all the tools in its deployment marker and updates them together.

• Use one project folder per book
• Re-run the setup after updating the toolkit to refresh the project files
• Use the check-only request first if you are unsure of the current state