README Writer

What this skill does

This skill directs the agent to read a project's structure, package.json (or equivalent manifest), source files, and any existing documentation, then generate a complete, professional README.md. The output adapts its tone and structure to the type of project — a CLI tool, a web app, a library, or a framework — and covers everything a new user or contributor needs to get started.

Use this when you have a new project without documentation, an outdated README, or an open-source project you want to present well.

How to use

Claude Code / Cline

Copy this file to .agents/skills/readme-writer/SKILL.md in your project root.

Then ask:

• "Use the README Writer skill to generate a README for this project."
• "Write a professional README.md for this repo using the README Writer skill."

The agent will explore the project structure on its own. You can also provide a description of what the project does if the code alone isn't self-explanatory.

Cursor

Add the "Prompt / Instructions" section to your .cursorrules file. Then open the relevant files (package.json, main entry point) and ask Cursor to write the README.

Codex

Paste your package.json (or equivalent), the main entry file, and a brief description of the project. Then include the instructions below.

The Prompt / Instructions for the Agent

When asked to write a README, follow these steps:

1. Gather information. Read:

• package.json, pyproject.toml, Cargo.toml, go.mod, or equivalent manifest for name, description, version, dependencies, and scripts
• The main entry point file to understand the project's purpose
• Any existing README.md, CHANGELOG.md, or docs/ folder
• The directory structure (top level) to understand the project layout
• LICENSE file to identify the license

1. Determine the project type and adapt accordingly:

• Library/Package — focus on API reference, installation, and code examples
• CLI tool — focus on installation, commands, flags, and usage examples
• Web application — focus on features, setup, and deployment
• Framework/boilerplate — focus on getting started, conventions, and architecture

1. Write the README with these sections (adapt as needed for project type):

• Title + Badges: Project name as an H1. Add relevant badges (build status, npm version, license) using standard badge formats. Only add badges for things that actually exist.
• One-line description: What the project does in one sentence.
• Features: 4–8 bullet points of the most important capabilities.
• Tech Stack: Brief list of core technologies used (language, framework, database, etc.).
• Prerequisites: What the user needs installed before they can run the project.
• Quickstart / Installation: The exact commands to install and run — copy-paste ready. Use code blocks with language tags.
• Usage: How to use the project after setup. For libraries: code examples. For CLIs: command syntax and examples. For apps: how to navigate key features.
• Configuration: Document environment variables, config files, or options. Use a table with columns: Variable, Default, Description.
• Project Structure: A concise directory tree showing where things live. Only include meaningful directories, not every file.
• API Reference (libraries/APIs only): Document public functions, methods, or endpoints with their signatures and descriptions.
• Contributing: How to contribute — fork, branch, PR process, code style, running tests.
• License: One line with the license name and a link to the LICENSE file.

1. Write in plain, direct language. Avoid filler phrases like "This project aims to..." — start with what it does.

1. Make all code blocks runnable. Every shell command must be accurate and complete. Test commands against the actual scripts in the manifest.

1. Do not invent features that don't exist in the code. Only document what is real.

Example

Input to Agent:

"Use the README Writer skill on this project. Here's the package.json: ``json { "name": "snap-resize", "version": "1.2.0", "description": "Fast image resizing CLI for batch processing", "bin": { "snap-resize": "./dist/cli.js" }, "scripts": { "build": "tsc", "test": "jest" }, "dependencies": { "sharp": "^0.33.0", "commander": "^12.0.0" } } ``"

Output from Agent:

# snap-resize

>

>

Fast image resizing CLI for batch processing — resize hundreds of images in seconds with a single command.

>

## Features - Batch resize entire directories of images - Output to JPEG, PNG, or WebP - Preserve or override aspect ratio - Set quality level per output format - Built on sharp for native-speed processing

>

## Prerequisites - Node.js 18 or higher - npm 9 or higher

>

## Installation ``bash npm install -g snap-resize ``

>

## Usage ```bash # Resize all images in a folder to 800px wide snap-resize ./photos --width 800 --out ./resized

>

# Convert to WebP at 80% quality snap-resize ./photos --format webp --quality 80 --out ./output ```

>

## Configuration | Option | Default | Description | |--------|---------|-------------| | --width | — | Output width in pixels | | --height | — | Output height in pixels | | --format | original | Output format: jpeg, png, webp | | --quality | 85 | Compression quality (1–100) | | --out | ./out | Output directory |

>

## Contributing 1. Fork the repo and create a branch: git checkout -b my-feature 2. Make your changes and add tests 3. Run npm test — all tests must pass 4. Open a pull request

>

## License MIT — see LICENSE