QingYunA/readme-skill

Latest recorded source activity: October 9, 2026 at 07:30

SkillsMP catalog refreshed: November 9, 2026 at 14:22

skills collected: 1

GitHub stars: 1
GitHub forks: 1

Install with an AI assistant

Copy this prompt into the AI assistant you're using.

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/QingYunA/readme-skill.

Repository README

Repository README source: README.md

Read the source README

Expand full README

<h1 align="center">readme-skill</h1> <p align="center"> <b>An agent skill for GitHub READMEs: write a new one, update it, slim it down, or review it.<br>Ships with a lint script that catches dead anchors, broken HTML and drift between language versions.</b> </p> <p align="center"> <a href="#install">Install</a> · <a href="#what-it-does">What it does</a> · <a href="#the-lint-script">Lint script</a> · <a href="skills/github-readme/SKILL.md">SKILL.md</a> · <a href="README.zh-CN.md">简体中文</a> </p> Once installed, ask your agent the way you normally would: ``` > Write a README for this repo > The README is too long, slim it down > We renamed --watch to --follow, update the README > Review the README before we go public ``` The skill picks the task, checks every claim against the code, and runs the lint script before it hands the README back. Here is the script on a real README from [answer-me-with-html](https://github.com/QingYunA/answer-me-with-html) at commit `f3082c9`, after a hand edit on GitHub: ``` README.zh-CN.md: 2 error(s), 1 warning(s) warn inflated-word:298 赋能 error broken-tag:9 </p has no ">" error unclosed-tag:7 <p> ``` ## Install You need Node.js 18 or newer for the lint script. The skill itself is plain Markdown. ```bash npx skills add QingYunA/readme-skill ``` It asks which agents to install into. The installer is [vercel-labs/skills](https://github.com/vercel-labs/skills). <details> <summary>Manual install</summary> Copy `skills/github-readme` into your agent's skill folder. For Claude Code: ```bash git clone --depth 1 https://github.com/QingYunA/readme-skill.git /tmp/readme-skill cp -R /tmp/readme-skill/skills/github-readme ~/.claude/skills/github-readme ``` </details> ## What it does | Task | When | What the agent does | | :--- | :--- | :--- | | New | No README yet, or a full rewrite | Lists the facts with their sources, asks you what only you know, writes, lints, then has a second agent check it | | Update | A feature changed | Marks each claim as still true, stale or unproven, and fixes it in place | | Slim | Over about 350 lines | Moves reference material to `docs/` and leaves a one-line summary with a deep link | | Review | Before a release | Reports problems and leaves the files alone | - **Facts come from the code:** commands from `package.json` and `--help`, defaults from the code that reads them. A claim with no source is left out. - **It asks before it guesses:** measured numbers, the story behind the project and comparisons with other tools come from you. Anything still missing stays as a `TODO`, and the lint script fails on it. - **Plain writing:** one idea per sentence, no marketing adjectives, and every number comes with its test conditions. - **Templates for five kinds of project:** agent skills and plugins, self-hosted apps, desktop apps, CLIs, and libraries. The skill text is in Chinese. It works the same way when the README you ask for is in English. ## The lint script ```bash node skills/github-readme/scripts/check_readme.mjs README.md README.zh-CN.md ``` No dependencies. It exits with 1 when there are errors. | Level | Check | | :--- | :--- | | error | Anchors that go nowhere, including `docs/*.md#anchor` and Chinese headings | | error | Relative files and images that do not exist | | error | Machine-local paths such as `/Users/...` or `file:///...` | | error | Leftover `TODO` and `{{placeholders}}` | | error | Unclosed HTML tags, and tags with a missing `>` | | error | A license badge with no `LICENSE` file | | warning | More than 350 lines, and inflated words | | warning | `README.md` and `README.<lang>.md` differ in sections, images, code blocks or badges | ## Background The rules come from the README of [answer-me-with-html](https://github.com/QingYunA/answer-me-with-html): more than 50 commits, two passes that cut it from 388 to 320 and from 361 to 310 lines, and a few fixes the maintainer made by hand on GitHub. Ideas from other README skills are in too: the fact list, the lint checks and the second-agent review come from [Omer-Dahan/readme-skill](https://github.com/Omer-Dahan/readme-skill), the task split from [crafting-effective-readmes](https://github.com/softaworks/agent-toolkit/tree/main/skills/crafting-effective-readmes), and the rule to ask instead of guess from [portfolio-readme](https://github.com/barbaria888/Claude-repo-writing-skill). ## Development ```bash npm test npm run lint:readme ``` ## License [MIT](LICENSE)

Loading the repository README…

Skills in this repository

classification pending

skill

occupation

description

updated

github-readme

unclassified

撰写、重构、瘦身或审查 GitHub 项目的 README.md（含双语 README、头部徽章、安装段、benchmark 主张、docs/ 分流）。当用户要求写 README、优化或美化 README、README 太长、改 README 头部或安装说明、或提到 github-readme 时使用。产出说人话、短句、数字可验证的工程师式 README。

Source text: Chinese

2026-10-09

Showing 1 of 1 collected skill.