lol, better name: the rest of the fucking owl.py
pick some file ending, .llm_prompt or something like that. ooooh better yet, combine that with a tag in the commit message, like #4LLM.
tag triggers a github workflow which globs all the cached .llm_prompt files and sends them to the completion endpoint with a "flesh out this code" prompt.
maybe you just want to have one function filled in. maybe you have a list of function names you want to have filled in.
maybe you have just a docstring and you want to have the whole file filled in.
to be really generic, could add some kind of in-document template, like a yaml header that contains the prompt(s)/examples/whatever.
when the CI is triggered, a new document is generated for each .llm_prompt. same name, different file extension. default to .py or .md
for extra customization, this system could start by reading a config file. e.g. change the file extension targeted, default file extension to write, llm api, etc.