Token导航 LogoToken导航

Agent Skill · 开发

technical-writing(技术写作)

用于辅助文档、README、Markdown、说明文和内容稿件的整理与改写。它适合让 Agent 提炼结构、补齐章节、统一术语、检查链接或把零散材料整理成可读文档。使用时应保留项目已有事实、命令和路径,不要把未确认的信息写成确定结论;涉及对外文案时,还需要控制语气,避免过度营销或夸大能力。

technical-writing 是开发类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证信息还不明确,第三方扫描已通过,按安装量在开发类的 21,746 个 Skill 里排第 2,897。

总安装 13,814周安装 361GitHub Stars 公开资料未说明下载 3,937核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:technical-writing(技术写作)
来源仓库:https://github.com/caidanw/skills
仓库路径:skills/technical-writing
安装命令:
npx skills add https://github.com/caidanw/skills --skill 'Technical Writing'
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/caidanw/skills --skill 'Technical Writing'

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

technical-writing 强调技术文档应言简意赅,每句话都有明确目的,去除冗余表达。适用于 README、API 说明与操作指南编写,提升开发者上手效率。提倡主动语态、命令式语气,如“运行 ./script.sh”而非被动描述。使用前请提供完整上下文与示例,避免断章取义导致误用。

涉及命令行操作时必须注明前提条件与预期结果,防止环境差异引发问题。

来源仓库内路径:skills/technical-writing

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/caidanw/skills --skill 'Technical Writing';命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

technical-writing 怎么安装?

在终端执行 npx skills add https://github.com/caidanw/skills --skill 'Technical Writing'。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

technical-writing 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

technical-writing 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

technical-writing 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

technical-writing 用的人多吗?

公开数据是累计安装 13,814 次,近一周安装 361 次,下载 3,937 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

technical-writing 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
technical-writing

技术写作 · 本页

13,814—接触敏感数据未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Technical Writing

Write for humans and machines. Every sentence earns its place or gets cut.

Core Principles

Lead with the point. First sentence = the takeaway. Context comes after, if needed.

One idea per unit. One point per sentence. One topic per paragraph. One concern per section.

Concrete over abstract. Name the thing. Show the example. Skip the preamble.

Active voice, imperative mood. "Add retry logic" not "Retry logic should be added" or "Added retry logic."

No filler words. Cut: "basically", "simply", "just", "in order to", "it should be noted that", "as mentioned above", "please note that", "going forward". If removing a word doesn't change the meaning, remove it.

No hedging unless uncertainty is the point. Say "This will break X" not "This could potentially have an impact on X."

Prefer short words. "use" over "utilize", "start" over "initialize", "show" over "indicate", "about" over "approximately."

Format: Git Commits

Structure:

<type>: <what changed>

<why it changed — optional, only if non-obvious>

Types: feat, fix, refactor, docs, test, chore, perf, ci

Rules:

  • Subject line under 72 characters
  • Imperative mood: "Add X" not "Added X" or "Adds X"
  • No period at end of subject
  • Body explains *why*, not *what* — the diff shows what
  • Reference issue numbers when relevant: Fixes #42

Good:

feat: add rate limiting to /api/upload

Prevents abuse from automated clients. Limits to 100 req/min
per API key. Returns 429 with Retry-After header.

Fixes #187

Bad:

Updated the upload endpoint to add some rate limiting functionality
so that we can prevent potential abuse issues going forward

Format: GitHub Issues

Bug Reports

## Problem
<What's broken — one sentence>

## Steps to Reproduce
1. <Exact steps>
2. <No ambiguity>

## Expected
<What should happen>

## Actual
<What happens instead>

## Context
- Version/commit: <hash or version>
- Environment: <OS, browser, runtime>
- Logs/screenshots: <if relevant>

Feature Requests

## Problem
<What user need is unmet — not the solution, the problem>

## Proposed Solution
<How to solve it — be specific>

## Alternatives Considered
<What else you evaluated and why it lost>

## Scope
<What's in, what's explicitly out>

Format: PRD (Product Requirements Document)

# <Feature Name>

## Problem
<Who has this problem. What they can't do. Why it matters now.>

## Solution
<What we're building. 2-3 sentences max.>

## Requirements

### Must Have
- <Requirement — testable, unambiguous>
- <Each one starts with a verb>

### Nice to Have
- <Lower priority items>

### Out of Scope
- <Explicitly excluded to prevent scope creep>

## Success Criteria
- <Measurable outcome>
- <How we know this worked>

## Technical Notes
<Constraints, dependencies, migration concerns — only if relevant>

## Open Questions
- <Unresolved decisions, with owners and deadlines>

Rules:

  • Requirements are testable. "Fast" is not a requirement. "Page loads in < 200ms at p95" is.
  • Each requirement starts with a verb: "Support", "Display", "Validate", "Allow".
  • Success criteria are measurable. If you can't measure it, sharpen it.
  • Open questions have owners. Unowned questions don't get answered.

Format: Technical Specs / Design Docs

# <Title>

## Context
<Why this doc exists. What decision or system it describes. 2-3 sentences.>

## Design

### Architecture
<How it works. Diagrams welcome, walls of text not.>

### Data Model
<Schema changes, new entities, relationships.>

### API
<Endpoints, contracts, examples.>

## Tradeoffs
<What you chose and what you gave up. Be honest about costs.>

## Risks
<What could go wrong. What's the mitigation.>

Format: Inline Code Comments

Only comment *why*, never *what*. The code says what.

Good:

// Retry 3x because the payment API returns transient 503s under load

Bad:

// Call the payment API with retries

Don't comment:

// Increment counter
count += 1

Format: PR Descriptions

## What
<One sentence — what this PR does>

## Why
<Motivation — link to issue if applicable>

## How
<Brief summary of approach — not a line-by-line walkthrough>

## Testing
<How you verified this works>

Anti-Patterns

The Wall of Context. Three paragraphs of background before the point. Invert it. Point first, context if needed.

The Passive Report. "It was determined that the service should be restarted." By whom? Say who did what.

Weasel Words. "Some users", "significant impact", "may cause issues." Quantify or be specific.

Premature Abstraction. Don't generalize when a concrete example communicates faster.

The Apology Prefix. Don't start with "Sorry for the long message" — make the message shorter instead.

Redundant Structure. Don't add sections just because the template has them. Empty "Risks: None" sections waste attention. Omit sections that add nothing.

Editing Checklist

Before publishing, read it once and cut:

  1. Sentences that repeat what another sentence already said
  2. Adjectives and adverbs that don't change the meaning
  3. Throat-clearing openers ("So basically...", "As you may know...")
  4. Sections with no actionable content
  5. Anything the reader already knows or can infer

平台分布

Codex

37.72%
按下载量换算1,485

Claude

28.82%
按下载量换算1,135

Cursor

18.6%
按下载量换算732

Gemini CLI

10.85%
按下载量换算427

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills