The claim check
This README is 28.6% accurate.
An AI wrote it. It looked perfect. Then verify_readme.py checked every claim against the repo: the wrong package manager, a script that doesn't exist, an env var nothing reads, a Docker service that isn't there, a docs page nobody wrote.
readme-forge won't finish until every ✗ is fixed or marked as a TODO.
$ python verify_readme.py README.bad.md --repo .✓ 6 command npm install [package.json exists]✗ 6 package_manager npm [repo uses pnpm (pnpm-lock.yaml), not npm]✗ 7 script pnpm run deploy [no 'deploy' script in package.json (scripts: dev, test)]✓ 7 package_manager pnpm [pnpm-lock.yaml matches pnpm]✗ 8 env_var REDIS_URL [not referenced in any code, config or .env file]✗ 9 command docker compose up worker [no service worker in docker-compose.yml (services: api, db)]✗ 12 path docs/setup.md [no docs/setup.md in the repo]claims: 7 checked · 2 verified · 5 not found · 0 unverifiable · 1 generic · accuracy 28.6%
The gallery
Everything it makes.
Every card is a real file in the repo. Open one for the command that makes it.
How it works
Facts first, then words.
- 01
Reads your repo
Manifests, lockfiles, scripts, env vars, Docker, CI and routes, across 15 ecosystems.
- 02
Writes from facts
Only what it can point to in a file, and a
TODOwhere it can't. - 03
Checks every claim
Commands, scripts, flags, env vars and paths against the code, in Claude Code or in CI.
Keep it true on every pull request with the GitHub Action:
- uses: tanmayai23/README-Editor-skill@v1