I have worked with some who believe everything it says, copy paste code without thought no matter how bad a solution it is, spend hours in pathological loops iterating in circles and getting nowhere, and various other ways to generate a lot of code that kind of works so long as you don't go beyond limits and constraints that the software engineer committing it has no understanding of.

I would definitely like some way to know if they can be trusted with an LLM or if someone is going to be hard reviewing everything they submit forever. Don't know quite how to test that but asking them to use an LLM must be part of it.