Interesting approach. How does this architecture handle the 'lost in the middle' phenomenon when the retrieval volume increases? I've found that simply increasing volume often degrades reasoning quality unless re-ranking is extremely aggressive.
we consider this a separate issue that should be addressed using special approaches. however, the proposed technique generally tends to decrease the probability of something being lost in the middle by minimizing the total size of the retrieved documents.
This is a useful collection. As someone building local agents, I find the biggest bottleneck isn't 'discovering' skills, but defining the permission boundaries for them. Does this directory categorize skills by their 'safety level' (e.g., read-only vs. write-access)?
This is such a great point. Security does become the bottleneck once you move past simple queries.
We haven't implemented safety categorization yet, but I'd love to explore it. Ideally, would you prefer to see flags like 'Filesystem: Read-Only' or 'Network: Blocked' directly on the skill card?
Thanks for the feedback!
