Built a defensive counterpart: check your own site the way REA would

[kriszwart]

Thanks for REA. It made something very clear to me: anything a frontend ships, an agent can now read and explain.

That made me want the reverse view for my own products, so I built Diarea (MIT): https://github.com/kriszwart/diarea

It scans a site you own (or your build folder) and reports what an agent like REA could learn from it: source maps, secrets in bundles, client-side logic, API maps, AI system prompts and so on. It scores the result and suggests fixes. I borrowed REA's Seen / Inferred / Unknown framing for findings, because it maps neatly onto the goal of pushing everything into Unknown on the server.

It only looks at pages you control, and it isn't meant as a way to block REA. It just helps builders see their own exposure first.

If you think it's useful, I'd be glad to keep the finding categories in line with REA's evidence model, or to add a "defending your own app" note if you'd ever want one. Happy to hear any feedback either way.