SkillAxe: Sharpening LLM-Authored Agent Skills Through Evaluation-Guided Self-Refinement
For developers of LLM-based agent systems, SkillAxe provides a fully unsupervised method to automatically improve the quality of skill documents, addressing a key bottleneck in deploying LLM agents.
LLMs struggle to write effective skill documents for agent frameworks. SkillAxe, an unsupervised self-refinement framework, improves LLM-authored skill pass rates by 28% relative on SkillsBench and closes 47-67% of the gap to human-authored skills, and on SpreadsheetBench raises pass rate from 16.0% to 52.0% using only 22 skills.
Skill documents, structured natural-language instructions that guide Large Language Model (LLM) agents, are critical to modern agent frameworks, yet LLMs struggle to write skills that actually work. On SkillsBench, human-authored skills improve pass rates by 16.2 percentage points, while LLM-authored skills provide no measurable gain. We introduce SkillAxe, a fully unsupervised framework that enables LLMs to iteratively diagnose and refine their own skills. SkillAxe decomposes skill quality into four interpretable dimensions (quality impact, trigger precision, instruction compliance with fault attribution, and solution-path coverage), producing structured improvement briefs that require no ground-truth labels, test suites, or environment rewards. On SkillsBench, SkillAxe improves pass rates by 28\% relative over unimproved LLM skills and closes 47--67\% of the gap to human-authored skills. We validate the approach as a continuous improvement engine in the wild on SpreadsheetBench, where a SkillAxe-built skill library learns from past agent trajectories and raises pass rate from 16.0\% to 52.0\% using only 22 skills.