ASD-STE100 Skill ZH
Skill that rewrites Chinese technical text into controlled language: one action per sentence, active voice, one term per idea.
- 97 ★ on the repository
- Last commit
Minimal access
Why this level
- The skill only rewrites text by rules, and the linter reads files and changes nothing in them.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/asd-ste100-skill-zhDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill only rewrites text by rules, and the linter reads files and changes nothing in them.
This is third-party code. Review the repository files before installing.
What it does
The skill turns vague, wordy Chinese technical text into controlled language: one sentence carries one action, verbs are active and one word keeps one meaning. Nominalizations, marketing adjectives and ritual phrases are removed. The reader is someone who cannot ask follow-up questions: another agent, a translation pipeline, a non-native reader or yourself half a year later. The rules are written anew for Chinese, since typical failures differ from English ones: empty verbs, mixed synonyms and semicolon chains. An instruction is capped at 25 characters and an explanation at 40. Modality such as maybe is left alone, since confidence is part of the content.
Who it is for. Authors of Chinese documentation and prompts whose texts are read by agents, translators and colleagues outside your team.
Good fit when
- An error message or tool description is ambiguous and an agent misreads it.
- A README reads heavily and needs an editor pass.
- Instructions for another agent must lose their ambiguity.
- A text must become unambiguous before it goes to translation.
Not a fit when
- You need to edit English text, for which the original ASD-STE100 skill exists.
- The tone of fiction matters more than unambiguity.
- The author style must stay, including hedges and cautious wording.
Example request
Rewrite this error message using controlled Chinese rules and show the violated rules in a table.Limitations
The skill supplies rules and before-after examples, and the model performs the check, so results depend on the agent. The bundled linter `scripts/ste-lint-zh.py` is mechanical only: it uses the Python 3.8+ standard library and catches formal violations such as sentence length and stock phrases. It cannot replace editing for meaning. By default the skill returns only the rewritten text, and a table of violations and original phrasing appears on request to show the diff. For Codex the rules go into `AGENTS.md`, with no separate skill install there.
How to disable. Remove the skill folder from the agent skills directory or the rules entry from `AGENTS.md`. Files the agent already rewrote stay in the project.
FAQ
How does it differ from the English ASD-STE100 skill?
It is not a translation: rules are rewritten for typical Chinese failures such as empty verbs, synonym rotation, semicolon chains and translationese. The list of differences is in NOTICE.md.
What sentence length do the rules set?
Up to 25 characters for instructions and up to 40 for explanations.
Does the skill change hedges like maybe?
No. Modality stays, because the confidence of the author is part of the content.
How do I see what exactly changed?
Ask for the diff or the violated rules. The skill returns a three-column table: rule, original and rewrite.
How do I use it with Codex?
Paste the rule block from the README into the project `AGENTS.md` or a global instructions file. The linter helps for a mechanical check.
Discussion
No questions or reviews yet.
Related
SkillsA skill that rewrites dense, ambiguous English into ASD-STE100 Simplified Technical English so another agent cannot misread it
A skill that rewrites technical text in ASD-STE100 Simplified English: short sentences, active voice, warnings before the step.
A skill with a controlled Chinese for technical writing: 40 rules, a banned-word vocabulary and a checker modeled on ASD-STE100.
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 makes an agent write docs and replies in plain English on ASD-STE100 rules: short sentences, one term per meaning, condition before command
Skill that translates English and French course handouts into Chinese and builds HTML, DOCX or Markdown with terminology.
README in short
The repository gives a working guide for rewriting Chinese texts, before-after comparisons and a dependency-free linter. Installation is described for Claude Code, Codex and DSH, where the skill is registered by a local plugin.
README badge
Original SKILL.md
--- name: ste-zh description: 把含糊、啰嗦、容易被误读的中文技术文本改写成受控中文（一句一动作、主动语态、一词一义、去名词化、去营销形容词与套话），用于工具描述、报错信息、agent 间指令、system prompt、状态汇报、README。含 Strict / 中文-顺 两档与确定性 linter。触发：受控中文、中文去歧义、STE100 中文化、把这段话写清楚、"让别的 agent 读不错"。不用于创意、营销、文学文案。 version: 1.0.0 --- # 受控中文技术写作（ASD-STE100 中文化） ASD-STE100 是航空业为了「维修工不可能读错」而做的受控英语：一个词只有一个意思，一句话只有一种结构。 本技能把同一条纪律搬到中文，读者换成**没有追问机会的另一方**：另一个 agent、翻译管线、非母语读者、 半年后的你自己。 中文的歧义来源和英语不同。英语主要栽在「一个词多个意思」和「一句话多种结构」上；中文主要栽在四件事： **虚动词**（进行 / 加以 / 作出 + 名词）、**套话**（需要注意的是、在一定程度上）、**同义轮换** （用户 / 客户 / 使用者 混用）、**一句串三件事**（用分号、破折号、「并且 / 同时 / 从而」串起来）。 所以下面的规则按中文重写，不是把英文规则逐条翻译过来。 规则分两类，本技能只能把其中一类做到「可核对」： - **结构规则**：讲句子形状，读完描述就能执行，可以放心硬性套用。 - **词汇规则**：需要一个受控词表才算标准。中文没有（也不该有）官方词表，所以词汇规则降级成 「方向」——挑最普通、最常见的说法，并在输出里说明这是方向而不是合规声明。 ## 什么时候用 - 一段中文说明、报错、日志、工具描述读起来含混、术语堆砌、或者绕。 - 文本要交给另一个 agent、翻译管线、非母语读者解析，**读错有代价**。 - 你在写 prompt、system message、工具描述，想在模型看到之前先消掉歧义。 - 你想要**前后对照**，看清违反了哪条规则、改完为什么更好。要就说 —— 默认输出只有改写后的正文。 不用在：创意文案、营销文案、文学、需要语气与节奏的地方。受控中文是故意平淡、字面化的。 ## 两档模式 改写前先选档。用户没说就按文本类型推断，并在输出里用一句话说明选了哪档。 **Strict（严格）** —— 程序步骤、报错信息、工具与函数描述、agent 间指令、安全相关文本。 凡是读错有代价的地方，全部规则执行，包括硬性句长上限和「一词一义」。 **中文-顺（STE-flavored）** —— README、PR 描述、变更日志、解释性文章。结构规则全执行， 词汇规则只作为方向：句长上限、主动语态、简单体貌、不用分号、去名词化、去营销形容词都保留， 但不锁死用词 —— 说明文需要一点弹性，全锁死读起来像换了个作者，而不是变清楚了。 两档和「结构 / 词汇」这条分界线是同一件事的两个方向：分界线说的是**哪些规则不依赖词表也能核对**， 两档说的是**对某类文本该执行到哪一步**。