I was under the impression the purpose of that is to give an LLM a cheap (in terms of context window) understanding of a repo, and not at all for the benefit of

I was under the impression the purpose of that is to give an LLM a cheap (in terms of context window) understanding of a repo, and not at all for the benefit of humans? Whether that's effective is another matter, even if the LLM generating the list does so correctly and updates the list consistently