Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

jl-content-rewriter

v1.0.1

提供高质量内容洗稿,保持原意,降低重复率,优化结构,去除AI痕迹,提升文章原创性和可读性。

0· 84·0 current·0 all-time
bypycoder@pyzxs

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for pyzxs/jl-content-rewriter.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "jl-content-rewriter" (pyzxs/jl-content-rewriter) from ClawHub.
Skill page: https://clawhub.ai/pyzxs/jl-content-rewriter
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install jl-content-rewriter

ClawHub CLI

Package manager switcher

npx clawhub@latest install jl-content-rewriter
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The name/description (content rewriter) aligns with reading an input document and writing a rewritten output. However the SKILL.md metadata declares a required env var JL_CONTENT_REWRITER_OUTPUT_DIR while registry metadata listed no required env — this is an inconsistency. The skill also requires writing output into the same directory as input by default, which is reasonable for a rewriter but should be explicit to users.
!
Instruction Scope
Runtime instructions direct the agent to read arbitrary file paths provided by the user and save outputs in the original file's directory; that is expected. Concerningly, several reference files instruct the skill to read other skills' SKILL.md files from ~/.openclaw/workspace/skills/... rather than invoking those skills via the platform API. Reading other skills' files on disk is scope creep (access to other skills' internals/configs) and unnecessary for the stated rewriting purpose. The workflow also explicitly aims to '去除AI痕迹' and '大幅降低重复率' (evade AI-detection and plagiarism checks), which raises ethical and compliance concerns even if technically coherent with the goal of producing novel wording.
Install Mechanism
Instruction-only skill with no install spec and no code files—lowest install risk. Nothing is downloaded or written by an installer step.
!
Credentials
The SKILL.md metadata requires JL_CONTENT_REWRITER_OUTPUT_DIR (and documents additional optional env vars), but the registry metadata earlier reported 'Required env vars: none' — a mismatch that should be resolved. The skill also implicitly expects filesystem access to input file paths and to the OpenClaw workspace (~/.openclaw/workspace/skills/...), which grants access beyond a single file and is not declared in registry metadata.
Persistence & Privilege
always:false and no install spec — the skill does not demand elevated persistence. However the instructions require writing output files into the input directory (potentially overwriting or adding files) and removing metadata from outputs; users should be aware of possible data loss or unexpected writes. The skill's guidance to read other skills' SKILL.md from the workspace touches other skills' data, which is a cross-skill access concern.
What to consider before installing
Before installing/using this skill, consider the following: - Resolve the metadata mismatch: SKILL.md declares JL_CONTENT_REWRITER_OUTPUT_DIR but the registry lists no required env vars. Confirm which env vars will actually be used and set them deliberately. - File access & overwrite risk: the skill reads arbitrary file paths you pass in and by default writes outputs into the same directory (naming {orig}_final.md). Use a safe test directory first, and back up originals before running to avoid accidental overwrite or data loss. - Cross-skill file reads: references instruct reading SKILL.md files under ~/.openclaw/workspace/skills/ (ai-humanizer, humanizer-zh, blog-rewriter, etc.). Ask the author why direct disk reads of other skills are needed instead of using platform skill-invocation APIs; this behavior grants access to other skills' files and may expose sensitive metadata. - Ethical/compliance risk: the workflow explicitly targets removing 'AI traces' and lowering similarity to evade detection. That can be used to bypass plagiarism detection or content-moderation systems; ensure you have the right to republish/transform any content you process and that use complies with platform policy and law. - Principle of least privilege: restrict the skill's allowed input paths to non-sensitive directories and avoid giving it access to system or credential files. If you proceed, test with benign sample files first and confirm exactly where outputs and temp files are written. If the author can (1) explicitly declare required env vars in registry metadata, (2) stop recommending direct reads of other skills' SKILL.md (or explain why and request explicit permission), and (3) document safeguards for overwrites and sensitive-data handling, the inconsistencies would be largely addressed.

Like a lobster shell, security has layers — review code before you run it.

latestvk978gbyvrdhxssrys0nmv3aefn857f7e
84downloads
0stars
2versions
Updated 1w ago
v1.0.1
MIT-0

⚙️ 环境变量配置

核心变量

变量名说明默认值
JL_CONTENT_REWRITER_OUTPUT_DIR输出目录输入文件目录
JL_CONTENT_REWRITER_FILE_SUFFIX文件名后缀_final.md
JL_CONTENT_REWRITER_KEEP_ORIGINAL_PATH保留路径结构false

配置示例

# 自定义输出目录
export JL_CONTENT_REWRITER_OUTPUT_DIR="/data/processed"

# 自定义文件名
export JL_CONTENT_REWRITER_FILE_SUFFIX="_final.md"

# 保留原始路径
export JL_CONTENT_REWRITER_KEEP_ORIGINAL_PATH=true

核心能力

输入原始文稿,输出改写后的全新文稿: ✅ 保持原文核心意思、信息点、篇幅长度不变(±10%篇幅误差) ✅ 100%替换原文措辞、句式结构、段落顺序,大幅降低重复率 ✅ 优化表达逻辑,修正病句,提升文案可读性 ✅ 支持自定义改写风格:正式/口语化/书面/活泼/严谨等 ✅ 去除AI味,AI文本转化为真实的人类文字。检测并消除内容 ✅ 写作风格及内容润色

工作流程(必遵守)

第一步:接收请求与文件处理

  1. 识别请求:当用户使用格式"技能jl-content-rewriter完成对下面文章的二次创作:[文件路径]"时激活
  2. 文件读取:读取指定文件内容,记录文件所在目录
  3. 目录确定:使用原始文件所在目录作为工作目录,确保输出文件与原始文件在同一目录

第二步:内容分析与洗稿

  1. 通读原文:提取核心信息点、逻辑结构、核心观点,统计原文篇幅(字数)
  2. 结构调整:打乱原文段落顺序,重新组织逻辑结构,不改变信息点先后的合理性
  3. 措辞替换:替换原文规则,请参考references/word-replace.md
  4. 逻辑优化:修正原文逻辑不通顺的地方,补充缺失的过渡句,提升可读性

第三步:质量校验

  1. 篇幅校验:最终输出篇幅和原文误差不超过±10%,不大幅删减原文信息点
  2. 查重校验:检查是否有连续重复内容,确保重复率低于30%

第四步:优化处理

  1. 去除AI味:调用去除AI味技能,具体执行请参考references/ai-humanizer.md
  2. 内容润色:调用写作风格润色技能具体执行请参考references/writing-polish.md

第五步:原创检查与相似度评估

  1. 原创性检查标准,请查看references/principled-review.md

  2. 相似度评估方法, 请查看references/evaluation-dimensions.md

  3. 平台抄袭风险评估表, 请查看references/risk-assessment.md

风险评估输出格式:

## 📊 相似度评估报告

### 总体相似度:XX%

### 各平台风险评估:
- ✅ **知乎**:安全(XX% < 25%)
- ⚠️ **微信公众号**:警告(XX% 在20-25%之间)
- ❌ **小红书**:高风险(XX% > 20%)
- ✅ **B站**:安全(XX% < 25%)
- ✅ **CSDN**:安全(XX% < 35%)
- ⚠️ **简书**:警告(XX% 在30-35%之间)

### 改进建议:
1. 重点优化【具体段落】降低表达相似度
2. 添加【具体类型】原创内容
3. 调整【具体部分】结构顺序

第六步:用户指令处理决策树

**15. 用户指令分析与响应流程:**请查看references/user-command.md

AI响应模板:

## 🔍 相似度评估完成

### 📊 评估结果
- 总体相似度:28%
- 原创性评分:7/10

### 🎯 平台适配情况
✅ **通过**:CSDN、B站
⚠️ **边缘**:知乎、简书
❌ **高风险**:微信公众号、小红书、抖音

### 💡 优化建议
1. **高优先级**:改写第三部分技术描述(当前相似度45%)
2. **中优先级**:添加个人使用案例
3. **低优先级**:调整文章开头表达

### ❓ 下一步操作
请选择:
1. **继续优化** - 重新洗稿降低相似度
2. **输出保存** - 直接保存当前版本
3. **指定平台** - 针对【平台名称】优化
4. **查看详情** - 显示详细评估报告

第七步:输出与保存

  1. 文件保存:在工作目录(原始文件所在目录)生成输出文件

  2. 命名规范:输出文件命名为{原始文件名}_final.md

  3. 最终收尾:对处理后的文案进行排版,去除所有中间过程标记,输出纯净的Markdown格式

目录管理原则

核心规则:

1、如果接收请求为文件地址,原始文案和最终输出在同一目录

2、终案终仅有处理后的文章内容,没有任何中间过程或元数据信息

用户文件:"/data/processed/douyin_7628062674416323849.txt
输出目录:"/data/processed

生成文件:
├── douyin_7628062674416323849.txt          (原始文件)
└── douyin_7628062674416323849_final.md     (洗稿后文件)

最终收尾与输出

文件输出规范

  1. 保存位置:输出文件必须保存在原始文件同一目录下
  2. 命名规则{原始文件名}_final.md
  3. 文件内容:纯净的、排版精良的Markdown格式,去除所有中间过程标记和元数据信息

AI响应格式

当技能执行完成时,AI应回复:

## ✅ 洗稿改写完成

### 📋 处理信息
- **原始文件**:{原始文件}或{原始内容}
- **输出文件**:{二次创作后新文件}
- **保存目录**:`{JL_CONTENT_REWRITER_OUTPUT_DIR}`
- **处理时间**:{当前时间}

### 📊 洗稿统计
- **原文字数**:{原文字数}
- **输出字数**:{输出字数}
- **篇幅变化**:{变化百分比}%(符合±10%要求)
- **重复率**:低于30%(通过查重校验)

### 👀 内容预览
{输出文件前5行内容预览}

### 🔧 处理流程
1. ✅ 通读原文并提取核心信息
2. ✅ 结构调整与段落重组
3. ✅ 措辞替换(同义替换>80%)
4. ✅ 逻辑优化与过渡补充
5. ✅ 篇幅校验(误差±10%以内)
6. ✅ 查重校验(重复率<30%)
7. ✅ 去除AI味处理
8. ✅ 内容润色优化
9. ✅ **专业排版处理**(新增)
10. ✅ 最终保存(去除所有元数据)

### 💡 使用说明
- 输出文件已保存在原始文件同一目录
- 文件为**纯净、排版精良**的Markdown格式,可直接使用
- 保留了原文所有核心信息点
- 语言通顺自然,符合日常生活化语气
- **不包含任何元数据信息**,只有文章内容

快速使用

针对文件二次创作

技能jl-content-rewriter完成对下面文章的二次创作:[文件路径]

示例

# 处理单个文件
技能jl-content-rewriter完成对下面文章的二次创作:/path/to/article.txt

# 输出结果(默认)
原始文件:/path/to/article.txt
输出文件:/path/to/article_final.md

针对内容进行二次创作

技能jl-content-rewriter完成对下面文章的二次创作:[文件内容]

示例

# 处理内容
技能jl-content-rewriter完成对下面文章的二次创作:`#Hermes Agent使用体验分享

上期视频测试了herme agent的安全性和skill自主迭代能力。本期继续讲解herme agent的高级用法。`

# 输出结果(默认)
原始内容:文章内容前500字
输出文件:/path/to/article_final.md

具体示例: 原始文案请参考,example/original_article.txt, 二次创作保存内容,参考example/final_article.md

依赖说明

本技依赖以下技能

- 去AI味道: ai-humanizer, humanizer-zh
- 内容润色: blog-rewriter, chinese-writing-polish

注意事项

  1. 信息完整性:洗稿改写时不允许大幅删减原文的核心信息点,所有原文提到的重要内容必须保留
  2. 合规性:所有输出内容必须符合国家法律法规,不出现违规内容、敏感词、虚假宣传内容
  3. 通顺自然:改写后的内容必须通顺自然,不能为了替换措辞出现不通顺的句子
  4. 自定义需求:用户有特殊要求(比如指定风格、指定关键词必须保留)时,优先满足用户需求
  5. 目录管理:必须确保原始文件和输出文件在同一目录下,便于文件管理

提示:首次使用前,设置环境变量,输出目录和终稿前后缀进行设置

Comments

Loading comments...