I'm honestly hitting a wall with this Claude Code setup and it's driving me nuts. I have this massive Python backend for a delivery app we're launching in like ten days and our test coverage is basically zero right now. I'm torn between two ways of using the CLI tool. One way is just using the execute command to let it script out entire test files from scratch, but then I'm worried it's gonna hallucinate a bunch of mocks that dont actually exist in our repo. My other logic was just sticking to the interactive mode and having it iterate on existing broken tests, but that feels so slow when I have like fifty modules to cover. I'm on a super tight budget too since we're bootstrapping this and I dont want to burn through my API credits in one afternoon just running loops that go nowhere. Ive been trying to figure out if the test skill specifically is better for Pytest integration or if I should just use the general coding skill and point it at my files. It keeps trying to rewrite my conftest.py which is breaking everything. Does anyone know which specific skill set or workflow actually handles Python mocking without losing its mind? I need to get this done by next Friday or our lead dev is gonna kill me...