← Back to Skills Marketplace
cccpan

Chinese Sensitive Words

by CCCpan · GitHub ↗ · v1.0.5 · MIT-0
cross-platform ✓ Security Clean
251
Downloads
0
Stars
2
Active Installs
5
Versions
Install in OpenClaw
/install chinese-sensitive-words
Description
Chinese sensitive word detection and content compliance checker (中文敏感词/违禁词检测). Scan text for banned, restricted, and risky words across Xiaohongshu (小红书), Do...
README (SKILL.md)

中文敏感词检测工具

检测中文文本中的敏感词/违禁词,支持小红书、抖音、快手、B站等主流平台。

快速开始

cd scripts/

# 检测文案中的敏感词
./check.sh "这是全网最好用的美白产品,效果立竿见影"

# 查询某个词的安全替换建议
./suggestions.sh "美白"

# 查询全部替换建议库
./suggestions.sh

功能特点

  • 海量词库 — 10万+词条,每日更新,覆盖政治、色情、暴力、赌博、毒品、广告法极限词、医疗功效词等
  • 多平台支持 — 小红书、抖音、快手、B站专属词库
  • 风险分级 — 🔴 高危(封号)/ 🟡 中危(限流)/ 🔵 低危(建议修改)/ 💡 提示
  • 替换建议 — 不只检测,还推荐安全替代词
  • 智能识别 — 谐音变体(薇信→微信)、跳字(加 微 信)、手机号/URL 检测
  • NER 过滤 — 智能过滤地名、人名、机构名,减少误报

命令参考

check.sh — 检测敏感词

检测文本中的敏感词/违禁词,返回风险等级和替换建议。

# 基本用法
./check.sh "要检测的文案内容"

# 关闭 NER 过滤(更严格的检测)
./check.sh "要检测的文案内容" --no-ner

# 从文件读取
./check.sh --file input.txt

参数说明:

  • 第一个参数:要检测的文本(最大 3000 字符)
  • --no-ner:关闭 NER 智能过滤,检测更严格
  • --file \x3C路径>:从文件读取待检测文本

输出示例:

⚠️ 检测到 4 个敏感词

风险概览: 🔴 高危=1 | 🟡 中危=2 | 🔵 低危=1

🔴 高危(可能导致封号/删帖)
  - "13812345678" — 分类: 手机号

🟡 中危(可能导致限流/降权)
  - "最好用" — 分类: 广告法极限词 → 建议替换: 很好用, 超好用
  - "美白" — 分类: 医疗功效 → 建议替换: 提亮, 焕亮

🔵 低危(建议修改)
  - "加微信" — 分类: 引流 → 建议替换: 私信咨询

suggestions.sh — 获取替换建议

查询敏感词的安全替换词。

# 查询指定词的替换建议
./suggestions.sh "美白"
./suggestions.sh "最好"

# 查询全部替换建议库(按分类展示)
./suggestions.sh

输出示例:

"美白" (医疗功效词替换)
建议替换: 提亮, 焕亮, 匀净, 透亮

配置

免费使用(每月 10 次)

无需任何配置,开箱即用。

无限使用(获取 Token)

前往 GitHub Issues 获取正式 Token。

获取后在 skill 目录下创建 .env 文件:

SENSITIVE_WORDS_TOKEN=your_token_here

或设置环境变量:

export SENSITIVE_WORDS_TOKEN=your_token_here

默认 API 服务

本工具默认使用以下 API 服务进行敏感词检测:

https://www.xdhdancer.top/api8888

该服务由本项目维护,仅用于敏感词检测,不会存储或分享您的文本内容。

自定义服务地址

如果使用私有部署的检测服务:

SENSITIVE_WORDS_API_BASE=https://your-server.com/api
SENSITIVE_WORDS_TOKEN=your_token_here

使用额度

类型 额度 说明
免费(无 Token) 10 次/月 开箱即用,无需注册
注册用户(有 Token) 无限制 获取 Token

支持的平台词库

平台 词库内容
通用 政治、色情、暴力、赌博、毒品、违法
小红书 广告法极限词、医疗功效、虚假宣传、焦虑营销、品牌词
抖音 直播违禁词、引流词、夸大宣传
快手 社区规范违禁词
B站 社区规范、内容审核词

风险等级说明

等级 影响 示例
🔴 高危 封号/删帖 政治敏感、色情、暴力、手机号
🟡 中危 限流/降权 广告法极限词、医疗功效、虚假宣传
🔵 低危 建议修改 引流词、促销词
💡 提示 注意措辞 焦虑营销、容貌身材相关

故障排除

"Rate limit exceeded" 错误:

  • 免费额度已用完(每月 10 次),明天会重置
  • 或前往 GitHub Issues 获取 Token,享受无限使用

"Connection timeout" 错误:

  • 检查网络连接
  • 如果使用自定义服务地址,确认地址可访问

检测结果不准确:

  • 尝试关闭 NER 过滤(--no-ner),获得更严格的检测结果
  • NER 过滤默认开启,会自动排除地名、人名等误报

许可证

MIT

Usage Guidance
This skill appears to be what it claims, but it transmits the text you check to a remote API by default. Before installing or using it: (1) Decide whether you are comfortable sending sensitive content to the default endpoint (https://www.xdhdancer.top/api8888). (2) Prefer obtaining a token from the listed source only if you trust the operator, and do not paste secrets into .env unless you control the machine. (3) If privacy is required, consider self-hosting and set SENSITIVE_WORDS_API_BASE to your own server, or inspect/host the server code before sending data. (4) Note minor metadata mismatches: the registry says no required env vars even though primaryEnv is SENSITIVE_WORDS_TOKEN and the scripts also accept SENSITIVE_WORDS_API_BASE — this is a documentation inconsistency, not a code-level red flag.
Capability Analysis
Type: OpenClaw Skill Name: chinese-sensitive-words Version: 1.0.5 The skill provides a utility for detecting sensitive Chinese words by sending text to an external API (xdhdancer.top). The scripts (check.sh and suggestions.sh) safely handle user input using jq to prevent shell injection and implement a local usage tracker in the user's home directory. While the tool exfiltrates the text content to a third-party server, this is the documented and intended behavior for the service, and no evidence of malicious intent, credential theft, or unauthorized execution was found.
Capability Assessment
Purpose & Capability
Name/description, required binaries (curl, jq), the primary credential (SENSITIVE_WORDS_TOKEN), and the included scripts all align with a networked sensitive-word detection tool. The scripts only reference expected items (.env for token/API base, and a local usage file).
Instruction Scope
The SKILL.md and scripts send the user-supplied text to a remote service (default https://www.xdhdancer.top/api8888) for analysis and will include the token as a Bearer header if provided. This is consistent with the stated purpose, but it means user content is transmitted off-device — the README claims the service "will not store or share your text" (a server-side claim you cannot verify locally).
Install Mechanism
No install spec is present (instruction-only plus shell scripts), so nothing is downloaded or installed by the skill bundle itself. This is the lowest-risk install posture.
Credentials
The only credential used is SENSITIVE_WORDS_TOKEN (declared as primaryEnv). However, registry metadata lists "Required env vars: none" which is inconsistent with primaryEnv being set. The scripts also honor an optional SENSITIVE_WORDS_API_BASE (not listed in metadata). These are small metadata mismatches but not themselves excessive: the requested token is proportional to the service's functionality.
Persistence & Privilege
Skill does not request elevated or persistent platform privileges. It writes a small usage file to $HOME/.sensitive-words-usage to track free-call quotas and reads a .env in the parent directory for configuration — both are limited, local persistence. always:true is not set and the skill does not modify other skills or system-wide config.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install chinese-sensitive-words
  3. After installation, invoke the skill by name or use /chinese-sensitive-words
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.5
- Updated free and unlimited usage instructions: Token acquisition now via GitHub Issues instead of WeChat. - Added documentation of the default API endpoint for sensitive word detection. - Clarified that the default API service does not store or share user text content. - Updated documentation links and wording to reflect the new token acquisition method and service information.
v1.0.4
- Contact method for unlimited token updated from GitHub Issues to WeChat (chenganp) in documentation and description. - References and instructions relating to GitHub Issues replaced with the new WeChat contact. - No logic, API, or feature changes; documentation update only.
v1.0.3
- Updated the instructions for obtaining a Token: now directs users to GitHub Issues instead of a WeChat contact. - Adjusted all relevant documentation to remove references to WeChat and include GitHub links for unlimited usage. - No changes to functionality or core features; documentation clarity improved.
v1.0.2
- Changed the free usage limit from 100 times/day to 10 times/month. - Updated instructions and tables to reflect new quota and token acquisition method (add WeChat: chenganp for unlimited use). - Clarified that formal token is now required for unlimited access. - No functional or script changes detected.
v1.0.0
Initial release: Chinese sensitive word detection and content compliance checker. 2 tools: check_sensitive_words and get_word_suggestions. Supports Xiaohongshu, Douyin, Kuaishou, Bilibili.
Metadata
Slug chinese-sensitive-words
Version 1.0.5
License MIT-0
All-time Installs 2
Active Installs 2
Total Versions 5
Frequently Asked Questions

What is Chinese Sensitive Words?

Chinese sensitive word detection and content compliance checker (中文敏感词/违禁词检测). Scan text for banned, restricted, and risky words across Xiaohongshu (小红书), Do... It is an AI Agent Skill for Claude Code / OpenClaw, with 251 downloads so far.

How do I install Chinese Sensitive Words?

Run "/install chinese-sensitive-words" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Chinese Sensitive Words free?

Yes, Chinese Sensitive Words is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Chinese Sensitive Words support?

Chinese Sensitive Words is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Chinese Sensitive Words?

It is built and maintained by CCCpan (@cccpan); the current version is v1.0.5.

💬 Comments