openai/codex-security ? reverse-engineered prompt

Build me a Node.js security tool called Codex Security that can scan a codebase for vulnerabilities, help validate findings, and support fixing them.

I want both a command line app and a TypeScript SDK. The CLI should let me sign in, run a scan on a folder, choose between a ChatGPT sign in or an API key, and compare two scans so I can see what is new, fixed, reopened, or still there. The scan results should save to a report file and keep a local history so future scans can be compared.

Make it work nicely for local use and also in CI, where I can pass an API key through environment variables instead of logging in. If the local state folder can’t be written to, let me point it somewhere else. Include a Docker setup for running repeatable scans too.

Please keep the API clean and easy to use from TypeScript, and make the whole thing feel polished and production ready. If you need current docs or best practices, look them up online.