SkillReducer: Optimizing LLM Agent Skills for Token Efficiency

2026-03-31

AI Summary

This work addresses the pervasive redundancy and non-essential content in large language model (LLM) agent skills, which leads to wasted context, diluted attention, and increased reasoning costs. The authors propose a two-stage optimization framework: first, adversarial incremental debugging is employed to compress or complete skill routing descriptions; second, a taxonomy-guided structural refactoring of skill bodies separates core rules from auxiliary content and incorporates an on-demand loading mechanism. This approach systematically uncovers inefficiencies in skill design, embodying the “less is more” paradigm. Evaluated across 600 skills and the SkillsBench benchmark, it achieves a 48% reduction in description length, a 39% compression of skill bodies, and a 2.8% improvement in functional quality, while maintaining an average retention rate of 0.965 across five diverse models, demonstrating strong generalization capability.

Abstract

LLM-based coding agents rely on \emph{skills}, pre-packaged instruction sets that extend agent capabilities, yet every token of skill content injected into the context window incurs both monetary cost and attention dilution. To understand the severity of this problem, we conduct a large-scale empirical study of 55,315 publicly available skills and find systemic inefficiencies: 26.4\% lack routing descriptions entirely, over 60\% of body content is non-actionable, and reference files can inject tens of thousands of tokens per invocation. Motivated by these findings, we present \textsc{SkillReducer}, a two-stage optimization framework. Stage~1 optimizes the routing layer by compressing verbose descriptions and generating missing ones via adversarial delta debugging. Stage~2 restructures skill bodies through taxonomy-driven classification and progressive disclosure, separating actionable core rules from supplementary content loaded on demand, validated by faithfulness checks and a self-correcting feedback loop. Evaluated on 600 skills and the SkillsBench benchmark, \textsc{SkillReducer} achieves 48\% description compression and 39\% body compression while improving functional quality by 2.8\%, revealing a \emph{less-is-more} effect where removing non-essential content reduces distraction in the context window. These benefits transfer across five models from four families with a mean retention of 0.965, and generalize to an independent agent framework.

Problem

• LLM agent
• skill optimization
• token efficiency
• context window
• redundancy

Innovation

• SkillReducer
• token efficiency
• adversarial delta debugging
• progressive disclosure
• LLM agent skills