How are teams using AI-enhanced flows for developing/maintaining/extending a non-trivial code base?

Best I can get is that the GitHub repo contains a "curated knowledge base" on what the code should do.

But it is unclear to me exactly how to keep these prompts in sync with the code base.

Are they like yacc source files, and every time we do a clean build of the repo, the generated code is blown away and AI uses them to re-generate the code again, this time according to, say, a few new features added?

Or do we periodically have AI read both the prompts and the code, and suggests edits from one to the other?

What are the best practices here? Are there GitHub repos which illustrate how its done in a world-class way?