SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories

2026-05-31

AI Summary

Existing training-free skill adaptation methods rely on full trajectories or conversation-level feedback, leading to coarse failure attribution, unstable updates, or excessive generalization. This work proposes the first training-free, step-level skill adaptation framework that, without fine-tuning the backbone model, explicitly identifies the first actionable erroneous step through trajectory analysis, precisely assigns responsibility to candidate skills, and performs targeted updates with acceptability validation. By establishing a direct link between step-level failure attribution and skill-level responsibility, the method enables a more stable and auditable skill maintenance mechanism. It consistently outperforms existing baselines across WebShop, PinchBench, and Claw-Eval, achieving up to a 1.7% absolute improvement in success rate and average score gains of 1.5 and 1.8 points, respectively.

Abstract

Large language model (LLM) agents increasingly rely on reusable external skills to solve long-horizon interactive tasks. Existing training-free skill adaptation pipelines usually update skills from full trajectories or session-level feedback, which makes failure attribution coarse and often produces unstable or overly broad revisions. We propose SkillAdaptor, a training-free step-level skill adaptation framework with explicit failure attribution, and it can plug into OpenClaw-class agent harnesses. Given a failed trajectory, SkillAdaptor identifies a first actionable fault step, links responsibility to candidate skills, and applies targeted updates under explicit acceptance checks while keeping the backbone frozen. We evaluate on WebShop, PinchBench, and Claw-Eval with Kimi-K2.5, GLM-5, and GPT-5.2. SkillAdaptor improves over no-skill and skill-adaptation baselines on all three suites, with the largest single-metric improvements of +1.5 points on PinchBench Avg Score%, +1.8 on Claw-Eval Avg Score, and +1.7 on WebShop success rate. These results indicate that step-level attribution supports more stable and auditable training-free skill maintenance\footnote{The code will be released at https://github.com/zjunlp/SkillAdaptor.}.

Problem

• skill adaptation
• failure attribution
• LLM agents
• trajectory analysis
• training-free

Innovation

• step-level failure attribution
• training-free adaptation
• LLM agents
• skill updating
• trajectory analysis