Identifies that the full source code (skill body) of a tool is the primary signal for LLM tool selection, far outweighing the importance of descriptions or metadata.
March 25, 2026
Original Paper
SkillRouter: Retrieve-and-Rerank Skill Selection for LLM Agents at Scale
arXiv · 2603.22455
The Takeaway
Most agent frameworks rely on short skill descriptions for retrieval to save context; this paper shows that approach causes up to 44% performance degradation. This shifts the design pattern for agentic tool-use toward indexing the implementation logic itself.
From the abstract
As LLM agent ecosystems grow, the number of available skills (tools, plugins) has reached tens of thousands, making it infeasible to inject all skills into an agent's context. This creates a need for skill routing -- retrieving the most relevant skills from a large pool given a user task. The problem is compounded by pervasive functional overlap in community skill repositories, where many skills share similar names and purposes yet differ in implementation details. Despite its practical importan