Trajectory-Guided Fault Localization for Agent Skill Evolution

Key points

• Agent skills provide reusable guidance for code agents, but incomplete or unsuitable guidance can impair task execution.1
• To reduce the manual effort of skill refinement, recent approaches use LLMs to generate revisions from execution feedback.1
• Specifically, SkillMorph compares failure and success evidence in abstracted trajectories across repeated runs and tasks, incorporating changes between evolution loops to identify suspicious actions.1
• 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.1

Sources (1)

• [1]Trajectory-Guided Fault Localization for Agent Skill Evolution
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 12:38 PM

  To address this gap, we propose SkillMorph, a skill-evolution approach based on trajectory-guided fault localization in agent skills.

  Agent skills provide reusable guidance for code agents, but incomplete or unsuitable guidance can impair task execution.

Extractive summary: sentences quoted from the sources.