would love to hear what Jake has to say about that !
its basically the Karpathy llm wiki concept , but generalized to a simple folder setup .
main use case is to set up a knowledge base that is both human and AI readable.
I can see it could be integrated with ICM with no issue and it involves pointing the agent\skill to specific types and other metadata , that will be present as YAML front-matter in the files, as well as index file in each folder.
one caveat to notice is that its completely open format so You have to think about the process of ingesting data into it , and ideally an agent should handle it .. or a single prompt if its not too complicated.
its still somewhat of a personal system , I think that if you develop tools that will be used by an organization then migrating to a Standardised database is probably the way to go , if you want to maintain order .. but I'm uncertain about that.