Back to Details

oh-story-claudecode-writing

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

oh-story-claudecode Writing Skill

oh-story-claudecode 写作Skill

Skill by ara.so — Daily 2026 Skills collection.
A comprehensive Chinese web novel (网文) writing skill pack for AI coding agents, covering the full pipeline: market scanning (扫榜), novel analysis (拆文), writing (写作), and AI-style removal (去AI味) for both long-form and short-form fiction.
该Skill由ara.so开发——属于Daily 2026 Skills合集。
这是一款面向AI编码Agent的综合性中文网文写作Skill包，覆盖长篇与短篇小说的全流程：扫榜（市场调研）、拆文（小说分析）、写作、去AI味（去除AI写作痕迹）。

What This Project Does

项目功能

oh-story-claudecode
provides structured Claude Code / OpenClaw skills that guide AI agents through every stage of professional web novel production:
  1. 扫榜 (Market Scan) — Analyze trending titles on 起点, 番茄, 晋江, 知乎盐言
  2. 拆文 (Novel Analysis) — Deconstruct golden chapters, pacing, hooks, emotional arcs
  3. 写作 (Writing) — Generate outlines, character sheets, chapter drafts with platform-specific style
  4. 去AI味 (De-slop) — Detect and remove AI writing artifacts, banned word lists, rewrite examples
oh-story-claudecode
提供结构化的Claude Code/OpenClaw Skill，引导AI Agent完成专业网文创作的每一个阶段：
  1. 扫榜（市场调研）——分析起点、番茄、晋江、知乎盐言平台的热门作品
  2. 拆文（小说分析）——拆解黄金章节、叙事节奏、钩子设计、情感弧线
  3. 写作——生成符合平台风格的大纲、角色卡、章节草稿
  4. 去AI味——检测并移除AI写作痕迹、禁用词汇，提供改写范例

Installation

安装方法

Via Claude Code / OpenClaw

通过Claude Code/OpenClaw安装

Feed the repo URL to Claude Code or OpenClaw and say "安装这个 skill":
https://github.com/worldwonderer/oh-story-claudecode
将仓库URL提供给Claude Code或OpenClaw，并发送指令“安装这个Skill”：
https://github.com/worldwonderer/oh-story-claudecode

Via CLI

通过CLI安装

bash
npx skills add worldwonderer/oh-story-claudecode -y
To update, re-run the same command.
bash
npx skills add worldwonderer/oh-story-claudecode -y
更新时，重新运行上述命令即可。

Available Skills & Triggers

可用Skill与触发方式

SkillSlash CommandNatural Language Triggers
story-long-write
/story-long-write
, 
/story
, 
/网文
"帮我开书", "写长篇", "搭大纲"
story-long-analyze
/story-long-analyze
"拆这本书", "分析黄金三章"
story-long-scan
/story-long-scan
"扫长篇榜单", "起点趋势"
story-short-write
/story-short-write
"写短篇", "写一个反转故事"
story-short-analyze
/story-short-analyze
"拆短篇结构", "分析情绪曲线"
story-short-scan
/story-short-scan
"知乎盐言风口", "短篇扫榜"
story-deslop
/story-deslop
, 
/去AI味
"这篇太AI了", "去掉AI感"
browser-cdp
/browser-cdp
"抓取榜单数据", "登录后爬取"
Skill名称斜杠命令自然语言触发词
story-long-write
/story-long-write
, 
/story
, 
/网文
"帮我开书", "写长篇", "搭大纲"
story-long-analyze
/story-long-analyze
"拆这本书", "分析黄金三章"
story-long-scan
/story-long-scan
"扫长篇榜单", "起点趋势"
story-short-write
/story-short-write
"写短篇", "写一个反转故事"
story-short-analyze
/story-short-analyze
"拆短篇结构", "分析情绪曲线"
story-short-scan
/story-short-scan
"知乎盐言风口", "短篇扫榜"
story-deslop
/story-deslop
, 
/去AI味
"这篇太AI了", "去掉AI感"
browser-cdp
/browser-cdp
"抓取榜单数据", "登录后爬取"

Project File Structure

项目文件结构

Writing content is managed on the filesystem, not accumulated in conversation context.
写作内容通过文件系统管理，不会堆积在对话语境中。

Long-form Novel (长篇)

长篇小说

{书名}/
├── 设定/
│   ├── 世界观/          # 背景设定、力量体系（按主题拆文件）
│   ├── 角色/            # 每个人物一个文件（沈栀.md、陆衍止.md）
│   └── 势力/            # 每个势力/组织一个文件（天机阁.md）
├── 大纲/
│   ├── 大纲.md          # 全书卷级结构
│   ├── 细纲_第001章.md  # 每章一个章纲，与正文一一对应
│   └── ...
├── 正文/
│   ├── 第001章_章名.md
│   └── ...
└── 笔记.md
{书名}/
├── 设定/
│   ├── 世界观/          # 背景设定、力量体系（按主题拆文件）
│   ├── 角色/            # 每个人物一个文件（沈栀.md、陆衍止.md）
│   └── 势力/            # 每个势力/组织一个文件（天机阁.md）
├── 大纲/
│   ├── 大纲.md          # 全书卷级结构
│   ├── 细纲_第001章.md  # 每章一个章纲，与正文一一对应
│   └── ...
├── 正文/
│   ├── 第001章_章名.md
│   └── ...
└── 笔记.md

Short-form Story (短篇)

短篇小说

{标题}/
├── 设定.md
├── 正文.md
└── 笔记.md
{标题}/
├── 设定.md
├── 正文.md
└── 笔记.md

Key Workflows

核心工作流

Workflow 1: Start a Long-form Novel from Scratch

工作流1：从零开始创作长篇小说

bash
undefined
bash
undefined

1. Scan the market for trending topics

1. 调研市场热门题材

/story-long-scan
/story-long-scan

2. Analyze a top-performing novel

2. 分析爆款小说

/story-long-analyze
/story-long-analyze

3. Begin writing with outline scaffolding

3. 基于大纲框架开始写作

/story-long-write

**Example agent interaction:**
User: 帮我开书，都市赘婿题材，起点男频
Agent triggers: story-long-write → Generates: 书名/设定/世界观/都市背景.md → Generates: 书名/设定/角色/主角.md → Generates: 书名/大纲/大纲.md (五步大纲法, 卷级结构) → Generates: 书名/大纲/细纲_第001章.md → Generates: 书名/正文/第001章_重生归来.md
undefined
/story-long-write

**Agent交互示例：**
User: 帮我开书，都市赘婿题材，起点男频
Agent triggers: story-long-write → Generates: 书名/设定/世界观/都市背景.md → Generates: 书名/设定/角色/主角.md → Generates: 书名/大纲/大纲.md (五步大纲法, 卷级结构) → Generates: 书名/大纲/细纲_第001章.md → Generates: 书名/正文/第001章_重生归来.md
undefined

Workflow 2: Analyze an Existing Novel

工作流2：分析已有小说

bash
/story-long-analyze
bash
/story-long-analyze

Paste chapter text or provide file path

粘贴章节文本或提供文件路径

Agent returns: 黄金三章分析, 爽点密度, 钩子类型, 节奏评分

Agent返回：黄金三章分析、爽点密度、钩子类型、节奏评分

undefined
undefined

Workflow 3: Write and Polish a Short Story

工作流3：创作并打磨短篇小说

bash
undefined
bash
undefined

Write

创作

/story-short-write
/story-short-write

Input: genre, emotional arc template, word count target

输入：题材、情感弧线模板、字数目标

Remove AI artifacts

去除AI写作痕迹

/story-deslop
/story-deslop

Input: draft text file path

输入：草稿文件路径

Output: rewritten version with AI-style markers removed

输出：移除AI写作痕迹的改写版本

undefined
undefined

Workflow 4: Scrape Market Data with Browser CDP

工作流4：通过Browser CDP抓取市场数据

bash
/browser-cdp
bash
/browser-cdp

Reuses existing browser login session via CDP protocol

通过CDP协议复用浏览器已登录会话

Scrapes 起点/番茄/晋江/知乎盐言 ranking pages without re-login

无需重新登录即可抓取起点/番茄/晋江/知乎盐言的榜单页面

undefined
undefined

Shell Script Examples

Shell脚本示例

Initialize a novel project directory

初始化小说项目目录

bash
#!/bin/bash
bash
#!/bin/bash

init_novel.sh - Bootstrap long-form novel file structure

init_novel.sh - 快速搭建长篇小说文件结构

BOOK_NAME="${1:-我的新书}"
mkdir -p "${BOOK_NAME}/设定/世界观" mkdir -p "${BOOK_NAME}/设定/角色" mkdir -p "${BOOK_NAME}/设定/势力" mkdir -p "${BOOK_NAME}/大纲" mkdir -p "${BOOK_NAME}/正文"
BOOK_NAME="${1:-我的新书}"
mkdir -p "${BOOK_NAME}/设定/世界观" mkdir -p "${BOOK_NAME}/设定/角色" mkdir -p "${BOOK_NAME}/设定/势力" mkdir -p "${BOOK_NAME}/大纲" mkdir -p "${BOOK_NAME}/正文"

Create starter files

创建初始文件

cat > "${BOOK_NAME}/笔记.md" << 'EOF'
cat > "${BOOK_NAME}/笔记.md" << 'EOF'

创作笔记

创作笔记

核心爽点

核心爽点

读者画像

读者画像

竞品参考

竞品参考

待办

待办

  • 完成世界观设定
  • 完成主角角色卡
  • 完成第一卷大纲 EOF
cat > "${BOOK_NAME}/大纲/大纲.md" << 'EOF'
  • 完成世界观设定
  • 完成主角角色卡
  • 完成第一卷大纲 EOF
cat > "${BOOK_NAME}/大纲/大纲.md" << 'EOF'

全书大纲

全书大纲

核心矛盾

核心矛盾

第一卷：（卷名）

第一卷：（卷名）

  • 起：
  • 承：
  • 转：
  • 合：
  • 起：
  • 承：
  • 转：
  • 合：

第二卷：（卷名）

第二卷：（卷名）

结局方向

结局方向

EOF
echo "✅ 项目目录已创建: ${BOOK_NAME}/" echo "📁 结构:" find "${BOOK_NAME}" -type f | sort
undefined
EOF
echo "✅ 项目目录已创建: ${BOOK_NAME}/" echo "📁 结构:" find "${BOOK_NAME}" -type f | sort
undefined

Batch process chapters for de-slop

批量处理章节去AI味

bash
#!/bin/bash
bash
#!/bin/bash

deslop_batch.sh - Run de-slop on all chapter files

deslop_batch.sh - 对所有章节文件执行去AI味处理

BOOK_DIR="${1:-.}" OUTPUT_DIR="${BOOK_DIR}/精修"
mkdir -p "${OUTPUT_DIR}"
for chapter in "${BOOK_DIR}/正文"/*.md; do filename=$(basename "${chapter}") echo "🔍 处理: ${filename}"

Invoke story-deslop skill on each file

Agent reads chapter, applies three-pass de-slop method, writes output

cp "${chapter}" "${OUTPUT_DIR}/${filename}" echo " → 输出: ${OUTPUT_DIR}/${filename}" done
echo "✅ 批量去AI味完成，共处理 $(ls ${BOOK_DIR}/正文/*.md | wc -l) 章"
undefined
BOOK_DIR="${1:-.}" OUTPUT_DIR="${BOOK_DIR}/精修"
mkdir -p "${OUTPUT_DIR}"
for chapter in "${BOOK_DIR}/正文"/*.md; do filename=$(basename "${chapter}") echo "🔍 处理: ${filename}"

调用story-deslop Skill处理每个文件

Agent读取章节内容，应用三遍去AI法，输出处理后的内容

cp "${chapter}" "${OUTPUT_DIR}/${filename}" echo " → 输出: ${OUTPUT_DIR}/${filename}" done
echo "✅ 批量去AI味完成，共处理 $(ls ${BOOK_DIR}/正文/*.md | wc -l) 章"
undefined

Check chapter word count targets

检查章节字数达标情况

bash
#!/bin/bash
bash
#!/bin/bash

wordcount.sh - Report word counts for all chapters

wordcount.sh - 统计所有章节的字数

BOOK_DIR="${1:-.}" TARGET="${2:-3000}" # Default target: 3000 chars per chapter
echo "📊 章节字数统计 (目标: ${TARGET} 字)" echo "================================"
total=0 for chapter in "${BOOK_DIR}/正文"/*.md; do filename=$(basename "${chapter}") count=$(wc -m < "${chapter}") total=$((total + count))
if [ "${count}" -lt "${TARGET}" ]; then echo "⚠️ ${filename}: ${count} 字 (不足目标)" else echo "✅ ${filename}: ${count} 字" fi done
echo "================================" echo "📖 总计: ${total} 字 | 平均: $((total / $(ls ${BOOK_DIR}/正文/*.md | wc -l))) 字/章"
undefined
BOOK_DIR="${1:-.}" TARGET="${2:-3000}" # 默认目标：每章3000字
echo "📊 章节字数统计 (目标: ${TARGET} 字)" echo "================================="
total=0 for chapter in "${BOOK_DIR}/正文"/*.md; do filename=$(basename "${chapter}") count=$(wc -m < "${chapter}") total=$((total + count))
if [ "${count}" -lt "${TARGET}" ]; then echo "⚠️ ${filename}: ${count} 字 (不足目标)" else echo "✅ ${filename}: ${count} 字" fi done
echo "=================================" echo "📖 总计: ${total} 字 | 平均: $((total / $(ls ${BOOK_DIR}/正文/*.md | wc -l))) 字/章"
undefined

Knowledge Base (References)

知识库（参考资料）

Each skill auto-loads relevant knowledge from 
references/
— only what's needed for the current task:
TopicContentSkill
大纲排布五步大纲法, 故事结构分级, 节点设计法, 升级感设计
long-write
人物设计角色设定, 人物提取, 关系映射, 动机链, 群像
long-write
钩子技法章尾钩子13式, 章首钩子7式, 段落级钩子, 悬念编排
long-write
情绪设计6种弧形模板, 期待感管理, 题材赛道策略
long-write
去AI味预防, 三遍去AI法, 改写范例库, 禁用词表
deslop
写作公式21大题材写作公式, 三翻四震, 感情线四阶段
short-write
女频写作女读者偏好, 情感描写, 感情线模式, 对标拆书
short-write
市场数据题材趋势, 平台特性, 投稿审核, 推荐安排
long-scan
/ 
short-scan
高级技法小纲四步法, 高潮逆推, 双线结构, AB交织法
long-write
风格模块对话, 打斗, 智斗, 镜头式写作, 装逼打脸, 白描
long-write
每个Skill会自动从
references/
目录加载当前任务所需的相关知识库内容：
主题内容对应Skill
大纲排布五步大纲法, 故事结构分级, 节点设计法, 升级感设计
long-write
人物设计角色设定, 人物提取, 关系映射, 动机链, 群像
long-write
钩子技法章尾钩子13式, 章首钩子7式, 段落级钩子, 悬念编排
long-write
情绪设计6种弧形模板, 期待感管理, 题材赛道策略
long-write
去AI味预防, 三遍去AI法, 改写范例库, 禁用词表
deslop
写作公式21大题材写作公式, 三翻四震, 感情线四阶段
short-write
女频写作女读者偏好, 情感描写, 感情线模式, 对标拆书
short-write
市场数据题材趋势, 平台特性, 投稿审核, 推荐安排
long-scan
/ 
short-scan
高级技法小纲四步法, 高潮逆推, 双线结构, AB交织法
long-write
风格模块对话, 打斗, 智斗, 镜头式写作, 装逼打脸, 白描
long-write

Supported Platforms

支持的平台

TypePlatforms
长篇起点中文网, 番茄小说, 晋江文学城, 七猫小说, 刺猬猫
短篇知乎盐言故事, 番茄短篇, 七猫短篇
类型平台
长篇起点中文网, 番茄小说, 晋江文学城, 七猫小说, 刺猬猫
短篇知乎盐言故事, 番茄短篇, 七猫短篇

Common Patterns

常用模式

Pattern 1: Market-to-Draft Pipeline

模式1：从市场调研到草稿的流程

/story-long-scan     → identify trending genre + tropes
/story-long-analyze  → paste top 3 chapters of a competitor novel
/story-long-write    → generate outline aligned with market gaps
/story-long-scan     → 识别热门题材与套路
/story-long-analyze  → 粘贴竞品小说的前三章内容
/story-long-write    → 结合市场空白生成大纲

Pattern 2: Revision Loop

模式2：修订循环

Draft chapter → /story-deslop → human review → /story-deslop (second pass)
章节草稿 → /story-deslop → 人工审核 → /story-deslop（二次处理）

Pattern 3: Skip Prep, Jump to Writing

模式3：跳过准备环节，直接写作

If you already have a direction, skip scanning and analysis:
User: 我已经有设定了，帮我直接写第一章

Agent: triggers story-long-write with --skip-prep flag
→ Reads existing 设定/ files
→ Generates 大纲/细纲_第001章.md
→ Generates 正文/第001章_*.md
如果已有创作方向，可跳过调研与分析环节：
User: 我已经有设定了，帮我直接写第一章

Agent: triggers story-long-write with --skip-prep flag
→ Reads existing 设定/ files
→ Generates 大纲/细纲_第001章.md
→ Generates 正文/第001章_*.md

Pattern 4: Short Story with Emotional Arc

模式4：带情感弧线的短篇小说创作

/story-short-write
→ Select arc template (e.g., "跌入深谷后反弹" / "平稳中突转")
→ Agent scaffolds: 设定.md → 正文.md (with embedded hook + reversal)
→ /story-deslop on 正文.md
/story-short-write
→ 选择情感弧线模板（例如："跌入深谷后反弹" / "平稳中突转"）
→ Agent搭建框架：设定.md → 正文.md（包含钩子与反转设计）
→ 对正文.md执行/story-deslop处理

Troubleshooting

故障排查

Skill not triggering

Skill无法触发

  • Ensure install completed: 
    npx skills add worldwonderer/oh-story-claudecode -y
  • Try explicit slash command: 
    /story-long-write
    instead of natural language
  • Re-run install to pull latest version
  • 确认安装完成：运行
    npx skills add worldwonderer/oh-story-claudecode -y
  • 尝试使用明确的斜杠命令：例如用
    /story-long-write
    替代自然语言触发
  • 重新运行安装命令获取最新版本

Browser CDP not connecting

Browser CDP无法连接

  • Ensure Chrome/Chromium is running with remote debugging enabled:
    bash
    google-chrome --remote-debugging-port=9222 --user-data-dir=/tmp/chrome-debug
  • Check CDP endpoint: 
    curl http://localhost:9222/json/version
  • The skill reuses your existing login session — log into the platform manually first
  • 确保Chrome/Chromium已启用远程调试模式：
    bash
    google-chrome --remote-debugging-port=9222 --user-data-dir=/tmp/chrome-debug
  • 检查CDP端点：运行
    curl http://localhost:9222/json/version
  • 该Skill会复用浏览器已登录会话，请先手动登录目标平台

Output files not being created

输出文件未生成

  • Verify the agent has write permissions to the working directory
  • Check that the book name doesn't contain special characters that break directory creation
  • Use the init script to pre-create the directory structure
  • 确认Agent对工作目录有写入权限
  • 检查书名是否包含会导致目录创建失败的特殊字符
  • 使用初始化脚本预先创建目录结构

AI artifacts still present after de-slop

去AI味后仍存在AI写作痕迹

  • Run 
    /story-deslop
    a second time (三遍去AI法 = three-pass method)
  • Provide specific examples of AI-sounding phrases for the agent to target
  • Check 
    references/禁用词表
    is loaded — mention it explicitly if needed
  • 再次运行
    /story-deslop
    （三遍去AI法即三次处理流程）
  • 提供具体的AI风格语句示例，让Agent针对性处理
  • 确认
    references/禁用词表
    已加载，必要时可明确提及该文件

Chapter too short / pacing issues

章节过短/节奏问题

  • Use 
    /story-long-analyze
    on your own draft to get density scores
  • Request "扩写" (expand) with specific scene types: "扩写打斗场景至3000字"
  • Reference the 节奏分析 knowledge module explicitly
  • 对自己的草稿运行
    /story-long-analyze
    获取密度评分
  • 请求针对性扩写：例如“扩写打斗场景至3000字”
  • 明确提及节奏分析知识库模块

Full Pipeline Example

完整流程示例

bash
undefined
bash
undefined

Step 1: Initialize project

步骤1：初始化项目

bash init_novel.sh "天机阁传说" cd "天机阁传说"
bash init_novel.sh "天机阁传说" cd "天机阁传说"

Step 2: Market research (agent scans 起点 male fantasy rankings)

步骤2：市场调研（Agent扫描起点男频玄幻榜单）

/story-long-scan → saves market notes to 笔记.md

/story-long-scan → 将市场调研笔记保存至笔记.md

Step 3: Analyze a competitor (paste 3 chapters into agent)

步骤3：分析竞品（向Agent粘贴3章内容）

/story-long-analyze → appends findings to 笔记.md

/story-long-analyze → 将分析结果追加至笔记.md

Step 4: Generate full outline + first chapter

步骤4：生成完整大纲+第一章

/story-long-write → creates 设定/, 大纲/, 正文/第001章.md

/story-long-write → 创建设定/、大纲/、正文/第001章.md

Step 5: Continue chapter by chapter

步骤5：逐章创作

/story-long-write → reads existing files, generates next chapter

/story-long-write → 读取已有文件，生成下一章内容

Step 6: Polish

步骤6：打磨内容

/story-deslop → rewrites 正文/*.md removing AI artifacts

/story-deslop → 重写正文/*.md文件，移除AI写作痕迹

Step 7: Word count check

步骤7：字数检查

bash wordcount.sh . 3000
undefined
bash wordcount.sh . 3000
undefined