Generating manifests for the test runner

Is there any documentation on how to use the manifest.py generator for
the WPT  test runner?

(Other than just reading through 600 lines of almost completely
uncommented code?  I'm a python newbie.)

I'm creating a new set of tests in a new directory and want to try
them in the runner, but when I try to run the tool I get "Not writing
manifest because working directory is not clean."  It's not an error
that's really helping me out...

Can I not run the tool until I've committed my tests? Until they've
been pushed to my remote fork? Merged to the upstream master?

Do I need to write a manifest for my test cases?

Is it just generated automatically for all the tests?  How do I
indicate what kind of tests they are?

 It's not obvious to me where to start here, in part because I've
already run the tool so I don't know what is generated vs. what is
baseline, and my previous questions indicated that people had checked
in things that really should've been generated... etc.

thanks,

Brad