SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories
Summary
SkillAdaptor is a training-free step-level skill adaptation framework with explicit failure attribution for LLM agents, improving performance on WebShop, PinchBench, and Claw-Eval.
View Cached Full Text
Cached at: 06/02/26, 03:23 AM
Paper page - SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories
Source: https://huggingface.co/papers/2606.01311
Abstract
Step-level skill adaptation framework with explicit failure attribution improves training-free skill maintenance for LLM agents in interactive tasks.
Large language model (LLM) agents increasingly rely onreusable external skillsto solve long-horizon interactive tasks. Existingtraining-free skill adaptationpipelines usually update skills from full trajectories or session-level feedback, which makesfailure attributioncoarse and often produces unstable or overly broad revisions. We propose SkillAdaptor, a training-free step-level skill adaptation framework with explicitfailure 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 onWebShop,PinchBench, andClaw-EvalwithKimi-K2.5,GLM-5, andGPT-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 onPinchBenchAvg Score%, +1.8 onClaw-EvalAvg Score, and +1.7 onWebShopsuccess rate. These results indicate that step-level attribution supports more stable and auditable training-free skill maintenanceThe code will be released at https://github.com/zjunlp/SkillAdaptor..
View arXiv pageView PDFAdd to collection
Get this paper in your agent:
hf papers read 2606\.01311
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2606.01311 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2606.01311 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2606.01311 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
SkillMaster: Toward Autonomous Skill Mastery in LLM Agents
This paper introduces SkillMaster, a training framework that enables LLM agents to autonomously create, refine, and select skills through trajectory-informed review and counterfactual utility evaluation.
Skill is Not One-Size-Fits-All: Model-Aware Skill Alignment for LLM Agents
This paper proposes MASA, a framework that adapts skills to each LLM backbone without modifying weights, using hierarchical evolution and a model-conditioned rewriter, achieving gains of up to 25.8 points over baselines.
SkillCAT: Contrastive Assessment and Topology-Aware Skill Self-Evolution for LLM Agents
SkillCAT is a training-free framework for LLM agent skill self-evolution that addresses limitations of single-trace bias, unverified merging, and full corpus loading via three stages: Contrastive Causal Extraction, Assessment-Augmented Evolution, and Topology-Aware Task Execution, achieving up to 40.40% improvement on benchmarks.
SkillLens: Adaptive Multi-Granularity Skill Reuse for Cost-Efficient LLM Agents
This paper introduces SkillLens, a hierarchical framework for adaptive multi-granularity skill reuse in LLM agents, demonstrating improved accuracy and cost-efficiency on benchmark tasks.
SkillAudit: Ground-Truth-Free Skill Evolution via Paired Trajectory Auditing
SkillAudit introduces a framework for evolving LLM agent skills without ground-truth feedback by using paired trajectory auditing and contrastive evaluation. It achieves 73.9% average task reward across 89 tasks, outperforming baseline methods.