README Generator
Drafts a clear, honest README from a project's actual files — what it does, how to install and run it, and how to contribute — without inventing features.
What this skill does
Key features
- Reads package manifest and entry files to describe the real project
- Produces the sections readers actually need, in order
- Includes copy-pasteable install and run commands from the real scripts
- Never invents features, badges, or benchmarks
Use cases
- Bootstrapping a README for a new project
- Refreshing a stale or thin README
- Standardising docs across many repos
SKILL.md
The skill definition — its metadata and the exact instructions an agent follows. Copy it into a SKILL.md file in your agent's skills folder.
| name | readme-generator |
|---|---|
| description | Draft a project README from its actual files, without inventing features. |
| allowed-tools | Read, Grep, Bash(ls:*) |
| honesty | describes only what exists |
Musts
- Read the package manifest (
package.json,pyproject.toml, etc.) and main entry files before writing. - Describe only what the project actually does — never invent features, integrations, or metrics.
- Pull install and run commands from the real scripts/manifest, and format them as copy-pasteable code blocks.
- Mark anything you can’t determine as an explicit
TODO, rather than guessing.
Guidelines
Default section order (drop any that don’t apply):
- Project name + one-line description
- What it does (2–4 sentences)
- Install
- Usage / quick start
- Configuration (env vars, if any)
- Contributing
- License
Tone
Plain, direct, and honest. A good README respects the reader’s time and never oversells.
FAQ
Will it add badges and benchmarks?
Only if they exist in the project. It won't fabricate CI badges, star counts, or performance numbers — inventing those is a common README anti-pattern.
What if the project is unusual?
It asks for or infers the entry point and scripts. If it can't determine how to run the project, it says so and leaves a clearly-marked TODO rather than guessing.
Before running any skill, read its instructions and the tools it's allowed to use, and test it on a safe target first. See LLM security best practices.