A README skill that deletes claims it cannot source

Description and use

The skill removes unsourced claims from README files, replacing them with verifiable facts and marking unverified features, improving documentation accuracy.

Problem

README files often contain unsourced claims such as install commands for non‑existent packages, configuration keys that the code does not use, or roadmap features described as if already released.

Outcome

The skill eliminates unsourced claims, replaces them with verifiable information, and flags unverified features, producing a more accurate and user‑friendly README.

Workflow

• 1. Build a fact map from the repository (package manifests, entry points, example scripts, tests, CI config, existing docs).
• 2. Generate README content only from supported facts.
• 3. Remove any claim that cannot be sourced rather than rewording it.
• 4. Record unverified items so they appear as limitations.
• 5. Apply a structure template that matches the project’s domain (e.g., CLI, desktop app, script) based on a study of 200 high‑star repositories.
• 6. Output the rewritten README.
• 7. Optionally run the skill via the CLI: npx skills add chromoany/readme-atlas and then Use $readme-atlas to rewrite the README.

Evidence

Before writing a sentence, the skill builds a fact map from the repository: package manifests, entry points, example scripts, tests, CI config, and any existing docs.

Source

When something cannot be sourced, the claim is removed rather than reworded.

Source

The 200-repository study is about structure, not quality. It tells you whether a README's shape matches what readers of that kind of project expect, not whether the project is worth using.

Source

The skill is plain Markdown — instructions plus reference files loaded on demand. No runtime, no API key.

Source