Progressive Discovery of CONTEXT.md

[DavidSchuldenfrei]

In a significant size project, CONTEXT.md can grow to being quite large.. This violates the idea of Progressive Discovery.
Suggestion: Have CONTEXT.md follow Progressive Discovery rules (a list of links to other files).
However the current format is very convenient for humans.
I suggest having 2 files, one (set of files) for agents with Progressive Discovery, and one single file for Human users

[blmille1]

@DavidSchuldenfrei I'm sorry if I'm misunderstanding you, but have you looked into CONTEXT-MAP.md? It's my understanding that it solves exactly that case, though it is described as being helpful for a monorepo. But since CONTEXT-MAP.md is only for glossary terms (AFAIK), I'd think it'd still be okay if you use it for your progressive disclosure.

[blmille1]

I'm also curious to hear the case for separate human vs. agent context models. Would the two files have duplicate content? If so, I discovered yesterday that duplication is a major source of agents not following your instructions, though that was in the case of AGENTS.md, which is different, but probably the same?