Pre-Flight Auditor for Chinese Social Media That Reads Context
It breaks the loop where boring Chinese social-media posts get no traffic and spicy ones get banned.

What it does
yuwen-publish-precheck is an agent skill that audits Chinese social-media copy—Douyin, Xiaohongshu, WeChat Channels—before you hit publish. It flags lines that violate platform rules or advertising law, cites the exact official source, and offers direct replacements that keep the tone but drop the violation. Everything runs locally; your drafts and accumulated rule library stay in a local data/ folder and never leave your machine.
The interesting bit Most “sensitive word” tools are dumb regex lists. This one runs a fast lexical scan for candidates, then lets an LLM judge whether the usage actually violates rules in context—so “lost eight pounds in two weeks” gets flagged as a medical claim for a diet product but might pass in a personal diary. It also debunks creator folklore, such as the myth that you must write “money” as “m” or “rice”; the README notes that homophones can actually trigger extra scrutiny.
Key highlights
- 72 official citations with verifiable links and dates; every flag is traceable to platform community conventions or Chinese advertising law.
- Calibrated on 38 real-world samples that actually passed platform review, so it avoids crying wolf over harmless hyperbole.
- 41 scene-aware detection modes that apply medical or financial rules only when relevant, rather than carpet-bombing every draft with every rule.
- Personal rule library: one-sentence feedback (“stop flagging ‘fat-burn’ in my fitness posts”) gets persisted locally and applied automatically next run.
- Explicitly refuses to teach evasion tricks like homophones, split characters, or obfuscation—treating those as risk amplifiers, not loopholes.
Caveats
- It does not promise 100% approval; final platform decisions depend on opaque internal models and account history.
- Currently covers only text (scripts, captions, titles); image and live-stream checks are on the roadmap but not implemented.
- Platform coverage is limited to Douyin, Xiaohongshu, and WeChat Channels; Kuaishou and others are planned.
Verdict Chinese content creators, MCN editors, and social-media marketers who want to stop guessing at the red lines should look here. If you don’t publish to Chinese platforms or you are looking for a guaranteed bypass, this won’t help—and the authors will tell you so upfront.
Frequently asked
- What is yuwen-cool/yuwen-publish-precheck?
- It breaks the loop where boring Chinese social-media posts get no traffic and spicy ones get banned.
- Is yuwen-publish-precheck open source?
- Yes — yuwen-cool/yuwen-publish-precheck is open source, released under the MIT license.
- What language is yuwen-publish-precheck written in?
- yuwen-cool/yuwen-publish-precheck is primarily written in Python.
- How popular is yuwen-publish-precheck?
- yuwen-cool/yuwen-publish-precheck has 620 stars on GitHub.
- Where can I find yuwen-publish-precheck?
- yuwen-cool/yuwen-publish-precheck is on GitHub at https://github.com/yuwen-cool/yuwen-publish-precheck.