yuwen-publish-precheck skill
A skill that checks a text before posting to Douyin, Xiaohongshu and WeChat Channels: what breaks the rules, which rule it rests on, and how to fix it
- 798 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Runs the local scan.py script and writes the profile, rules and case reviews into the data folder inside the skill folder
- Reads your drafts and notes on account restrictions, so work material can end up in the data files
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/yuwen-publish-precheckDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs the local scan.py script and writes the profile, rules and case reviews into the data folder inside the skill folder
- Reads your drafts and notes on account restrictions, so work material can end up in the data files
This is third-party code. Review the repository files before installing.
What it does
yuwen-publish-precheck checks scripts, articles, notes, subtitles and cover text before posting to the Chinese services Douyin, Xiaohongshu and WeChat Channels. The agent passes the material through two layers. First the `scripts/scan.py` script, plain Python 3.10 with no third-party dependencies, finds candidate words by rules and your own lists. Then the agent decides from context: the same phrase passes in a personal story and breaks the rules next to a course link. The verdict is given per platform: can post, post after edits, or do not post. Each objection comes with the place in the text, a rule number and a link to the official source, and for fixes the skill offers phrases you can drop in right away that keep the author's voice. After the edit it rechecks the text and gives a pre-posting checklist for the platform. The rules live in `references/`, forty-two of them, each tied to sources: the advertising law, regulator documents, platform and community rules, with a verification date. Context is handled by conditions: advertising norms apply only to commercial material, and medical and financial ones are added when the text touches those topics. Your trigger words, exceptions and cases with penalties are saved in the `data/` folder on your computer, and on request the skill also analyzes reduced reach.
Who it is for. For authors, marketers and teams who run accounts on Douyin, Xiaohongshu and WeChat Channels and want to know about problems before the platform notifies them.
Good fit when
- A script for a video with sales or advertising needs checking for prohibited wording before posting to Douyin or Xiaohongshu
- You need to rewrite a text so that it keeps its hook and meaning but stays within the platform's rules
- A video was throttled or removed and you need to work out the likely causes and fix the lesson in personal rules
- You have your own list of banned words and want to plug it into the check
Not a fit when
- The content goes out on foreign platforms such as X or YouTube: the author excludes them from the scope
- You need an account growth strategy or text written from scratch: the skill checks and edits finished material
- The material is in Russian: the candidate-word lists and rules are built for Chinese texts and Chinese law
- You need a guarantee that the platform will let the material through: the skill catches violations visible before posting and does not predict the platform's model or account weight
Example request
Use yuwen-publish-precheck and check this script for posting on Douyin: the text is belowLimitations
Version 1.0.6 of 2026-09-16, MIT license. Quotes from laws, regulator documents and platform rules do not receive a new license by being included in the repository.
The skill's instructions, reference files and word lists are written in Chinese and rest on Chinese norms, so a Russian-speaking reader will have to translate the sources. The working area is limited to three platforms, and others are not checked.
The author does not promise moderation will pass and gives no masking tricks: no homophone swaps, splitting into parts, replacement with symbols or covering text, even when the user asks. The skill also does not predict the platforms' internal models. The project description says 38 scripts that platforms let through average 0.8 candidate words per text, which does not guarantee the false positive rate on your topic.
The data you save in `data/` stays local and does not go to Git.
How to disable. Delete the skill folder (for example `~/.cursor/skills/yuwen-publish-precheck`) from your agent's skills directory. Personal rules live in `data/` inside that folder, so save them first if you need them.
FAQ
How is this check different from a banned-word list?
A word in the text is only a signal, and the decision depends on context. A phrase about income in a story about your own experience passes and breaks the rules next to a course sale. The word list finds candidates and the agent makes the decision.
Which platforms does the skill know?
Douyin, Xiaohongshu and WeChat Channels. Each has its own rules file in `references/`, and the verdict is built per platform.
How does the skill know a text is commercial?
It looks at the profile in `data/profile.md` and asks once if there is nothing there. With no answer the material is treated as commercial and checked more strictly, because missed advertising is worse than one extra remark.
How do I add my own words and exceptions?
Tell the agent in a phrase: watch for a word, treat a word as acceptable in your topic, add a working formulation. Entries go to `data/my-rules.md` and `data/expressions.md`, and big word lists go to `data/wordlists/` one word per line.
What if a video was already throttled?
Tell the agent: the skill opens `references/diagnose.md`, works through four types of causes with examples, saves the case and turns the finding into a rule. It does not invent violations when there is no platform notice.
Discussion
No questions or reviews yet.
Related
SkillsHuman Writing
human-writing
A skill for lively Chinese writing: helps the agent write and edit so a specific person seems to speak, not a model
qu-ai-wei (去 AI 味)
qu-ai-wei
A skill that edits Simplified Chinese: cuts stock phrases and calques while keeping facts, numbers and the author's voice
Content skills: article illustrations, covers, infographics, comics, translation, URL to Markdown and social posting
Ten skills run social media content from brief and strategy through platform copy and short video to publishing and review
An OpenClaw skill that runs a Xiaohongshu account through the built-in browser: feed analysis, ideas, note drafting, comment replies and a knowledge base
A WeChat article layout skill: turns Markdown into paste-ready HTML with inlined styles that survives pasting into the official account editor
README in short
The project description in Chinese opens with a flow diagram and a real example: a source script with a sale, the check result and the corrected text. Then come tables of the project's strengths, a section on accumulating personal rules, installation, first phrases for the agent and test details. The repository holds reference files per platform and industry, templates, evaluation cases, rule integrity tests and a changelog.
README badge
Original SKILL.md
--- name: yuwen-publish-precheck description: 国内自媒体发布前风险自审与保意修复：审口播稿、文章、图文笔记、字幕、封面文字能不能发抖音/小红书/微信视频号，给出具体位置、依据和可直接替换的改稿；被限流/删除/处罚后帮你复盘归因；你的行业敏感词、误报白名单和踩坑案例会沉淀成个人规则库，越用越准。Use when 用户说"能不能发""审一下稿子""查违禁词/敏感词""会不会限流/被限流了""帮我改成能发的版本""发布前检查""被平台处罚/删除了""帮我盯着这个词""导入违禁词表"。Not for 海外平台(X/YouTube)内容审核、起号涨粉策略、写稿创作本身。 --- # Publish Precheck｜发布前审 审核的目的是让内容安全地发出去，不是把稿子改成没人看的官样文章。词面命中只是线索，违规与否取决于语境、商业属性和主张对象；修复必须保住作者的口吻、钩子和信息量，只动真正越线的部分。 ## Outcome Contract - **Outcome**：对一份待发布内容给出逐平台结论（可发/改后可发/不建议发）、带原文位置和规则依据的问题清单、可直接替换的修复稿，以及发布前检查单。 - **Done when**：词面预检已跑、候选逐条判定完、（改稿时）修复稿复检通过、报告已按输出模板给出并存档。 - **Evidence**：每个"必改"项都引用原文位置和规则编号；规则文档中每条规则附官方来源引文与链接。