I start with “Is this worth automating at this level?”. So many times I have people asking for advice on UI tests they are writing that are just bad automated tests. Probably bad tests in general.
Assuming I’m writing an appropriate test: make sure all my branches and environment is setup, and get something mostly working (dirty if needs be). Once I’ve got the bones of what I’m doing I’ll sort out models, data, comments etc. Work out if there other useful tests, or exploratory test ideas while I’m here.
Lots of useful commits, linting, and doing the teams pre Merge Request checklist and we’re ready for review.