Guide · Agent skill
How to use the Long-form novel analyzer agent skill (story-long-analyze)
Updated · written with AI from the project's own documentation and checked against our catalogue data
story-long-analyze is an agent skill that teaches an AI agent to act as a structure analyst for long serialised novels, mostly Chinese web fiction. It builds a chapter index, summarises each chapter, then writes a readable report on plot causality, timelines, character relationships, emotional hooks, pacing, settings and writing style. After this guide you will know what the skill does, how to install it, how to ask for an analysis in natural language, and what to check before you start.
Long-form novel analyzerzenstory-ai/oh-story-claudecode · MIT Download the skill Quick start
- 1Download and unzip. Download the skill ZIP from the directory page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
- 2Or upload in the Claude apps. In the Claude apps, open Settings, choose Capabilities, then Skills, and upload the skill.
- 3Confirm Python 3. Make sure python3, python or py -3 runs on your machine, since the bundled scripts need it.
- 4Prepare the novel file. Save the novel as a plain .txt file, ideally UTF-8. GBK files are converted automatically.
- 5Ask in plain language. Say 帮我拆这本书 or give the path to the .txt file, or use /long-form trigger /长篇拆文.
- 6Review the opening chapters. When the agent pauses after chapters 1 to 3, read the quick preview and decide whether to continue.
- 7Read the main report. When the analysis finishes, open 拆文报告.md first and use the other folders for detail.
What the skill makes the agent do
An agent skill is a folder of instructions and optional scripts that an AI agent loads when a task matches it. This skill is designed for one job: reading a full novel and producing a structured analysis of it.
The skill works in stages. It first builds a chapter index with a bundled script, then reads the opening three chapters closely and writes a quick preview. Later stages extract a summary for each chapter, then cover plot and emotional rhythm, characters and settings, and a main report. A style profile is produced last. Results are saved as files in a folder named after the book, so work can resume later and earlier results can be reused.
- Golden three chapters: a deep breakdown of the opening chapters, plus a quick preview file
- Per-chapter summaries and a chapter index
- Plot, emotion, rhythm, character relationships and settings
- A main report and a style profile
Install the skill (ZIP download or Claude apps)
Download the ZIP of the skill from the directory page and unzip it. For Claude Code, put the unzipped folder into ~/.claude/skills/ to make it available everywhere, or into a project's .claude/skills/ folder to limit it to that project.
If you use the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill there instead.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/
- Claude apps: Settings → Capabilities → Skills, then upload
Check what you need before you start
The skill runs bundled Python scripts, so Python 3 must be available. The scripts can be launched as python3, python or py -3; the agent uses whichever works on your machine. You also need the novel as a plain-text file. UTF-8 is preferred; if the file is GBK, the skill converts it to UTF-8 first.
Running sub-agents for chapter extraction is recommended. If your agent cannot spawn sub-agents, the skill falls back to a single agent. The skill does not use the network.
- Python 3 (python3, python or py -3)
- A plain-text .txt novel file, ideally UTF-8
- Sub-agent support recommended; single-agent fallback otherwise
- Novels under about 20,000 characters are better suited to a short-form skill
Trigger the skill with a natural request
You can invoke the skill with the slash commands /story-long-analyze or /长篇拆文, or simply ask in plain language. The skill responds to requests such as 帮我拆这本书, 拆这本书, 分析黄金三章, 深度拆解 and 完整拆解, and also to providing a path to a novel text file.
When you give a file path, the agent saves the text as the book's source file inside a project folder named after the book, then starts with the index and opening chapters. It will pause after the first three chapters and ask whether to continue, so you can decide how much of the book to analyse.
- /story-long-analyze or /长篇拆文
- 帮我拆这本书 / 拆这本书
- 分析黄金三章
- 深度拆解 / 完整拆解
- Provide the path to the novel's .txt file
What the output folder contains
The main entry point is 拆文报告.md, the main report. Chapter-level deep breakdowns of chapters 1 to 3 sit in a chapters folder, alongside per-chapter summaries. Plot causality, pacing, emotional modules and scattered plot points are kept in a plot folder, characters and settings in their own folders, and the writing style in 文风.md.
Because the work is saved to files, you can stop and resume. The agent reads the saved progress record rather than starting over.
- 拆文报告.md: the main report to read first
- 章节/: golden-three deep breakdowns and per-chapter summaries
- 剧情/: plot, rhythm and emotion modules
- 角色/, 设定/, 人物关系图/: characters, settings and relationship map
- 文风.md: writing style notes
Safety notes: scripts, local files and copyright
The skill runs scripts on your machine and reads and stores the whole manuscript locally. It does not use the network, according to its listing. Only process a novel if you have the legal right to process its full text.
The skill is written to transfer only abstract mechanisms. It is not meant to reproduce a book's specific characters, settings, signature scenes or original sentences, so the output should be used for structural lessons.
- Runs local Python scripts; review them if your policy requires it
- Reads and stores the full text locally in the book's project folder
- Does not use the network
- Only use texts you have the right to process
Tips for better results
Start with the opening three chapters. The quick preview answers most questions about hooks and first impressions, and you can decide afterwards whether the full analysis is worth the time.
If an analysis stops because the file changed or chapter numbers do not line up, the agent will explain the choice and suggest a default. Keep the source file in a stable location and avoid editing it mid-analysis.
- Use UTF-8 text when you can to avoid garbled output
- Ask for the opening three chapters first, then decide whether to continue
- Ask for the style profile separately if you only need writing notes
- Keep the project folder in place so earlier results are reused
What people use it for
- Studying how a successful web novel hooks readers in its first three chapters
- Mapping the pacing and emotional payoffs of a long serial across many chapters
- Building chapter-by-chapter notes for a narrative structure course
- Extracting abstract structural lessons to inform your own serial plan
- Reviewing a writing style profile separately from the full analysis
Frequently asked questions
Is the story-long-analyze skill free to use?
The skill is published under the MIT licence, as stated on its listing. Check the licence file in the source repository for the full terms before redistributing it.
How do I install story-long-analyze?
Download the ZIP from the directory page and unzip it into ~/.claude/skills/ for personal use or a project's .claude/skills/ folder for one project. In the Claude apps, upload it under Settings, Capabilities, Skills.
What do I need to run it?
You need Python 3 available as python3, python or py -3, and a plain-text novel file, ideally UTF-8. The skill does not use the network, and it works best when the agent can spawn sub-agents.
Does it work on short stories?
Not well. The skill suggests a different short-form skill for texts under about 20,000 characters or without recognisable chapter headings.
Can it analyse non-Chinese novels?
The skill is described for long Chinese web fiction, and its trigger phrases are in Chinese. Its source does not state support for other languages, so test on a short sample first.
Will it copy a book's characters or scenes into my writing?
No, the skill is written to transfer only abstract mechanisms and avoid reproducing specific characters, settings, signature scenes or original sentences.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use story-long-analyze if you want a structured, chapter-by-chapter study of a long novel and you have the right to process its full text. Pick a different tool if you only need to analyse a short story, a novel under about 20,000 characters, or want to copy a book's specific settings and scenes.
Download the skill