nit: If there is an error here, the messages won't specify that it comes from reading an output file vs running the executable directly right? Might be helpful to differentiate that for the user.

Also, if the file response is valid Json but unsuccessful what happens here? In that case we want to re-run the executable.

If there is an error here, the messages won't specify that it comes from reading an output file vs running the executable directly right?

That's correct. It should be easy to add that

if the file response is valid Json but unsuccessful what happens here?

Good question, I believe it throws an error same as if it's the response. Are you suggesting that if the file JSON contains success: false, we should silently retry the executable?

Is there a good document that lists the expected logic for the outputFile, because it's not very intuitive. I followed the NodeJS implementation, although I may have missed some things.

go/pluggable-auth-design but it doesn't look like that has a great flow diagram. But yes, the expected behavior is that if the output file is parseable (I.e. good JSON) but success:false, then we would re-run the executable instead of just returning the error again

I have manually run the tests in go/pluggable-auth-sdk-testing and they are now passing