DEV Community

arafatruetbd
arafatruetbd

Posted on

I Built an AI Skill That Audits Your Next.js README

Have you ever opened a project README and found outdated information?

Maybe it says the project uses Redis, but Redis is no longer used. Or it tells you to run npm, while the project actually uses pnpm.

This is a common problem because the code changes, but the README often doesn't.

So I built Next.js README Auditor.

GitHub Repository

What does it do?

It is an AI Agent Skill that checks a Next.js repository and creates or updates its README.md based on the actual code.

Instead of trusting the existing README, it checks things like:

  • Next.js router and project structure
  • Dependencies and their real usage
  • Package manager and scripts
  • Routes and Server Actions
  • Database and authentication usage
  • Environment variables
  • Tests and CI
  • Deployment configuration
  • Outdated README claims

The basic process is:

Inspect → Verify → Write → Validate

A simple example

Imagine a README says:

Built with Next.js, PostgreSQL, Redis, Docker, and NextAuth.
Enter fullscreen mode Exit fullscreen mode

But the code only actually uses:

Next.js
PostgreSQL
Docker
Enter fullscreen mode Exit fullscreen mode

The auditor removes the unsupported claims instead of blindly copying the old README.

That's the main idea: documentation should describe what the code actually does.

Security

The skill does not read real secret files such as .env or .env.local.

For environment variables, it uses safe templates such as .env.example.

It also ignores generated folders like node_modules and .next.

Why I built it

I wanted a simple way to make AI-generated README files more trustworthy.

AI is good at writing documentation, but writing a nice README is not enough.

The information also needs to be correct.

That's why this project focuses on evidence before documentation.

If you work with Next.js projects, you can try it here:

Next.js README Auditor on GitHub

Feedback and contributions are welcome.

Top comments (0)