Top AI Product

SkillOpt, from Microsoft Research, is the top-upvoted agent paper on Hugging Face today. It tackles a layer above individual skills: the executive strategy for a self-evolving agent — deciding which skills to create, refine, keep, or discard as the agent accumulates experience.

## The self-evolving-skills space

A growing body of work frames the same closed loop: for each new task, an agent selects relevant skills, uses them to guide execution, then updates its skill collection based on what happened. The open problem is governance — without a strategy, skill libraries bloat with redundant or low-value entries, and the agent slows down sifting through them. SkillOpt’s “executive” framing targets exactly that curation problem.

## Why it matters

Skills have become the dominant way to specialize general agents — this site has covered agent-skills, dotnet/skills, and scientific-agent-skills among others. The next question is automatic: which skills should an agent grow on its own, and which should it prune? A system that manages its own skill portfolio strategically, rather than just accumulating, is what turns “skills” from a static library into a learning loop.