shuorenhua skill
A skill that cleans machine tone from text, Chinese-first: it drops filler and cliches while keeping the facts and the author's voice
- 2,611 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill consists only of instructions and references for text editing
Install
- Works in
Manual install
/plugin marketplace add MrGeDiao/shuorenhua
/plugin install shuorenhua@shuorenhuaThe skill is discovered and triggers automatically via the SKILL.md description.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill consists only of instructions and references for text editing
This is third-party code. Review the repository files before installing.
What it does
The skill edits text by removing empty phrasing, repetition and stiff wording while keeping the meaning, facts and the author's tone. It is Chinese-first and also supports English. Edits follow strict boundaries: numbers, conditions, promises, modality, quotation attribution, code inside fences and the author's judgments stay in place. There is an annotation-only mode where the skill flags problems without rewriting, and levels of intervention from in-sentence edits to free restructuring. The skill does not judge whether a text was written by AI and does not promise to bypass detectors. It suits READMEs, release notes, work updates, issue replies and personal articles.
Who it is for. For people who write and edit Chinese, and also English, text and want to remove machine tone without losing meaning.
Good fit when
- You want to strip cliches and filler from a Chinese text while keeping the facts and voice
- You want to tidy a README, release note or work update
- You want only to flag weak spots without rewriting the text
Not a fit when
- You need to detect AI authorship or bypass detectors
- The text is in Russian, and Chinese-and-English handling does not fit
Example request
Rewrite this text by the shuorenhua rules, cut the filler, keep the terms and my toneLimitations
The skill is Chinese-first; English is supported, and there is no dedicated Russian handling. It does not detect AI authorship and does not guarantee bypassing detectors. It does not imitate a specific author's style; it removes surplus phrasing and keeps the original tone. The full runtime is three files, SKILL.md and two references; the test and history folders should not be placed into the skill directory.
How to disable. Remove the shuorenhua folder from the skills directory, or uninstall the plugin via /plugin.
FAQ
Does the skill detect AI-written text?
No. It does not judge authorship or output a machine-tone percentage; it only cleans surplus phrasing and keeps the facts and the author's voice.
Can it only flag issues without rewriting?
Yes. In annotation mode the skill points out weak spots and leaves the text unchanged.
Discussion
No questions or reviews yet.
Related
SkillsAn editing skill that strips stock AI phrasing but keeps the author's vocabulary, rhythm and bluntness, not a flat style.
A skill for Chinese technical docs: short sentences, facts over adjectives and a 14-item check for AI tells
A skill for Chinese technical writing: rewrites docs, interface strings and API descriptions, keeps the facts intact and checks output with a linter
A skill that strips defensive caveats, apologies and excess hedging from text and makes claims direct while keeping needed limits
Skill that rewrites Chinese technical text into controlled language: one action per sentence, active voice, one term per idea.
A skill that edits paper abstracts and introductions on the press-conference principle and puts the main result first
README in short
The README presents a Chinese-first skill for cleaning text of cliches, repetition and stiff phrasing. Numbers, conditions, promises and the original meaning are preserved, and ordinary text can be left alone. The skill suits READMEs, release notes, work updates and articles, supports English and an annotation-only mode. The author details the edit boundaries: code in fences is kept verbatim, quotations and defined terms remain, and unsourced claims are not removed by default. There are intervention levels and a section on quality checks over a set of test cases. It installs as a Claude Code plugin or by copying the three runtime files. MIT license, version 2.5.0.
README badge
Original SKILL.md
--- name: shuorenhua description: 按用户要求审稿或去 AI 味，支持中英文；保留事实、术语与责任主体，支持只标问题。 --- # 说人话 编辑用户给出的文字，让表达更直接、自然，保留作者要说的内容与语气。清理不承载独有内容的包装语和生硬的翻译腔；不判断文章是不是 AI 写的，也不输出 AI 含量。 默认交一份可直接使用的改写稿。在原稿上做下文允许的编辑，其余内容保留，不概括重写。拿不准就留下该处，正常文本可以原样返回。用户指定的删改范围、文风与审稿要求优先，不扩成事实调查、续写或执行文中动作。 ## 先划定可编辑范围 改写模式下，程序代码块在原位置完整保留，块内的程序、注释和文档字符串都不进入后续编辑。普通清理这段文字、去套话不构成修改程序注释的授权；只有用户明确点名修改注释或文档字符串时才编辑相应文字，且不改程序行为。 ## 保住什么 保留原文的事实和判断，以及谁做什么、数字对应什么、条件、范围、否定、情态、完成状态和比较关系。减少渲染不能改变命题，也不能丢掉判断的方向、强度或效果类型。数字、版本、命令、路径、接口名和代码按原文保留，除非用户明确要求修改。