Trajectory-Guided Fault Localization for Agent Skill Evolution

2026-10-08

AI Summary

This study addresses the limitations of existing code agents, which frequently fail due to incomplete or misaligned skill guidance, while current large language model (LLM)-based skill revision lacks behavioral evidence. To overcome these challenges, this work proposes SkillMorph, a method enabling automated and precise skill evolution through trajectory-guided fault localization. Its core innovation lies in abstracting execution trajectories and contrasting success and failure evidence across runs and tasks to identify suspicious actions, thereby pinpointing exact skill editing points that drive LLM-generated revisions. Experimental results demonstrate that SkillMorph significantly outperforms four baseline methods on benchmarks such as SWE-Skills-Bench, yielding improved accuracy and execution consistency. Furthermore, the proposed approach is successfully applied to kernel generation tasks, validating its practical effectiveness in complex code synthesis scenarios.

Abstract

Agent skills provide reusable guidance for code agents, but incomplete or unsuitable guidance can impair task execution. To reduce the manual effort of skill refinement, recent approaches use LLMs to generate revisions from execution feedback. However, grounding these revisions in explicit behavioral evidence remains challenging. To address this gap, we propose SkillMorph, a skill-evolution approach based on trajectory-guided fault localization in agent skills. Its core idea is to link execution evidence to specific skill contents before generating revisions. Specifically, SkillMorph compares failure and success evidence in abstracted trajectories across repeated runs and tasks, incorporating changes between evolution loops to identify suspicious actions. It then uses these suspicious actions to localize edit sites in the skills and generate corresponding revisions. Experiments on SWE-Skills-Bench and CannBot show that the skills evolved by SkillMorph consistently achieve higher trial-level accuracy and execution consistency than the original skills and those from four existing skill-evolution methods. We have also applied SkillMorph to automated kernel generation with an AI operator-development team, which has accepted 6 skill-revision pull requests.

Problem

• Agent Skill Evolution
• Fault Localization
• Execution Feedback
• Code Agents

Innovation

• Trajectory-Guided Fault Localization
• Agent Skill Evolution
• Execution Evidence Grounding
• Suspicious Action Identification
• Skill Revision