technical-writing skill
A calm tone, evidence before conclusions and a buzzword blocklist for internal technical writing in Chinese.
- 220 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill is a set of text rules: it runs nothing and makes no network calls, only the style of the texts the agent edits changes.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/technical-writingDetects 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 is a set of text rules: it runs nothing and makes no network calls, only the style of the texts the agent edits changes.
This is third-party code. Review the repository files before installing.
What it does
technical-writing edits internal technical texts in Chinese: design documents, review notes, postmortems and talks. The agent gets the tone of a technical report instead of advertising or an essay: observation first, then generalization, then judgment, with no hasty conclusions. The rules live in `SKILL.md`: the order of a long text (background, breakdown of the current state, hypothesis, section-by-section answers, return to the hypothesis, impact, observability and evaluation), corner quotes `「」` and no dashes or hyphens as clause separators, a buzzword blocklist such as `赋能`, `闭环` and `抓手` with replacement by a concrete action, eleven bad-and-good walkthroughs and an editing checklist. Phrases that only guide the reader through the text are cut, and evaluative words are replaced by constraints and numbers.
Who it is for. People who write internal documents in Chinese, such as engineers and tech leads, and want to strip out marketing tone and hasty conclusions.
Good fit when
- An internal design document or review is being written and the tone has to suit peers and a manager.
- An article draft is overloaded with style and its conclusions sit ahead of the evidence.
- Chat logs and discussion outcomes are turned into a document that can be forwarded.
- Module boundaries, architecture evolution, observability or an evaluation method are being explained.
- A manager needs a three-part TL;DR: cause, hypothesis, decision.
Not a fit when
- Interface text, button labels or marketing copy is needed: the author excluded those genres.
- The task is help for external users or a mechanical API reference.
- The text is in Russian or English: the rules rest on Chinese phrasing, quotes and a blocklist of Chinese jargon.
- A literary or journalistic style is wanted: the skill suppresses metaphors and vivid turns.
Example request
Rewrite this design document with the technical-writing skill: drop the evaluative words and put evidence before conclusions.Limitations
All rules and examples in `SKILL.md` are written for Chinese: dropping dashes and hyphens as separators, the `「」` quotes and the jargon blocklist do not carry over to Russian or English. The skill changes tone and structure but verifies no facts: the author of the text supplies codes, metrics and issue links. The file is large, about 600 lines, with eleven worked examples. For Codex the repository has `agents/openai.yaml`, and the skill has to be registered there following the Codex documentation. There are no scripts or a linter: checking follows the checklist inside the file.
How to disable. Delete the `technical-writing` folder or symlink from `~/.claude/skills/` or another agent's skills directory. Remove the mention in `CLAUDE.md` or `AGENTS.md` by hand if you added one.
FAQ
How do I install technical-writing in Claude Code?
Run `npx skills add luoling8192/technical-writing`: the installer detects the agent itself. Manually: `git clone https://github.com/luoling8192/technical-writing.git ~/repos/technical-writing` and a symlink at `~/.claude/skills/technical-writing`.
Which texts is it for?
Internal technical texts in Chinese: design documents, review notes, postmortems, architecture write-ups and talks.
Which agents are supported?
Claude Code, Cursor, Codex, GitHub Copilot, Windsurf, Gemini, Cline, AMP and others. For Codex the repository has `agents/openai.yaml`.
What does the blocklist forbid?
Words that mimic information without content: `生态`, `赋能`, `闭环`, `抓手`, `沉淀`, `落地`, plus evaluative `极致` and `丝滑`. In their place the skill asks for an action, an object or a metric.
What structure does a long text follow?
Background, a breakdown of the current state, a hypothesis, section-by-section answers that return to the hypothesis, impact on code and process, observability and evaluation, then a brief close.
Discussion
No questions or reviews yet.
Related
SkillsA 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 for Chinese official documents that picks the document type, builds the structure and strips hollow boilerplate.
A skill that edits paper abstracts and introductions on the press-conference principle and puts the main result first
A skill that cleans machine tone from text, Chinese-first: it drops filler and cliches while keeping the facts and the author's voice
58 skills for lawyers under Chinese law: advice, litigation, criminal defence, contracts, compliance and source checks.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| technical-writing (this card) | no | 220 | |
| Chinese Tech Doc Style | no | 1,574 | edits documentation, interface strings and API descriptions with a linter check; here internal design documents, reviews and postmortems |
| zh-tech-writing | no | 406 | removes 14 AI tells in Chinese documentation; here the order of argument, a jargon blocklist and the structure of a long document |
README in short
The author credits @nekomeowww for ideas and feedback. Installation through skills.sh covers the main agents, and the manual route symlinks the repository into `~/.claude/skills/`. The license is MIT.
README badge
Original SKILL.md
--- name: technical-writing description: Use when drafting or revising internal technical design docs, review notes, architecture writeups, or sharing posts in Chinese, especially when the tone should stay evidence-backed, calm, and suitable for peers or leaders rather than marketing or literary writing. --- # Internal Technical Writing ## Overview Use this skill when the writing target is an internal技术方案、评审稿、架构说明、分享文稿，重点在于把背景、猜想、证据、抽象、实现影响面讲清楚。 这类写作更接近技术汇报，不是宣传稿，也不该写成散文。语气要平稳，判断要有来路，段落要能顺着问题往下读。 ## When to Use - 在写内部技术方案、设计说明、评审文档、分享稿 - 在改一篇已经有内容、但文风过满、结论过快、证据不足的技术文章 - 在解释代码、模块边界、架构演进、可观测性、评测方法 - 在把聊天记录、讨论结论整理成可传播的技术文档 **When not to use** - UI 文案、按钮文案、产品营销稿 - 面向外部用户的帮助文档 - 纯粹的 API reference 或机械说明书 ## Core Stance 先接住读者，再往下解释。 常见写法是作者替读者抢先下判断，像 `很清楚`、`说明了`、`显然`、`真正`。这里更适合换成三步： 1. 我们观察到了什么 2. 基于这些观察，可以归纳出什么 3. 这些归纳暂时更适合落在哪类实现或组织方式里 如果要给出更强的判断，至少满足下面任一条件： - 前文已经提出了对应的猜想，这一段是在 callback - 同一段上下文里已经有交叉验证的 evidence - 已经引用了具体模块、代码、issue、trace、指标或实验结果 这些 bad case 往往有同一种共性： - 句子承担的是“主持文章”的功能，不是“推进论证”的功能