AGENTS.md gets it wrong in 2 ways

Agents.md is just wrong altogether. if you're using claude just let it summarize for you and save it in its own way. There is absolutely no reason to standardize because the agentic code may have different ways of picking stuff up anyway.

Too big to fit all this crazy stuff?

  Project overview
  Build and test commands
  Code style guidelines
  Testing instructions
  Security considerations

The only thing that might be useful is this:

ui is a CRA app, it uses yarn dev (even that is more than you need)

server is a Python app and it uses make

yeah everything else both claude and cursor easily find, build and run tests autonomously without even describing these projects.