mlsanityData to 7 Oct 2026

Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills

arXiv 2603.25158 ↗First posted 26 Mar 2026cs.AIcited by 161

On the map: LLM agents and tool use → Agent skill learning and evolution

Abstract (arXiv)

Large Language Model (LLM) agents increasingly rely on domain-specific skills, yet manually authoring such skills does not scale, and skills generated purely from parametric knowledge often miss critical operational pitfalls. We introduce Trace2Skill, a framework that consolidates broad execution trajectories in parallel into a unified skill directory through inductive reasoning over agent experience. Trace2Skill supports both deepening existing human-written skills and creating useful skills from weak LLM-generated drafts. Experiments demonstrate the effectiveness of Trace2Skill across diverse domains, including office workflows, math reasoning, and vision QA. Importantly, the evolved skills are not merely memorized artifacts of the trajectories used to create them: they often transfer across model scales, across model families, and to out-of-distribution settings. For example, skills evolved from Qwen3.5-35B trajectories improve a Qwen3.5-122B agent by up to 57.65 percentage points on WikiTableQuestions. Further analyses show that Trace2Skill outperforms sequential skill editing and ReasoningBank-style retrieval memories, compresses recurring failures and workarounds into standard operating procedures (SoPs), and yields portable skills that can be reused without parameter updates or test-time retrieval.

Key facts. Cited by 161 papers in the map (data through 7 Oct 2026). Placed in Agent skill learning and evolution, one of 16 directions in LLM agents and tool use. Builds on 26 of its 35 references that are in the map. First posted on arXiv 26 Mar 2026.

Builds on

26 of 35 references.

Closest ancestors · 5

Same direction, newest first

  1. 2026MetaClaw: Just Talk -- An Agent That Meta-Learns and Evolves in the Wild17 Mar
  2. 2026ARISE: Agent Reasoning with Intrinsic Skill Evolution in Hierarchical Reinforcement Learning17 Mar
  3. 2026EvoSkill: Automated Skill Discovery for Multi-Agent Systems3 Mar
  4. 2026Organizing, Orchestrating, and Benchmarking Agent Skills at Ecosystem Scale2 Mar
  5. 2026SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning9 Feb

Other references · 21

  1. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
  2. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  3. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
  4. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,288
  5. 2015Compositional Semantic Parsing on Semi-Structured Tablescited by 509
  6. 2024Agent Workflow Memorycited by 342
  7. 2025ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memorycited by 253
  8. 2025SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skillscited by 174
  9. 2021HiTab: A Hierarchical Table Dataset for Question Answering and Natural Language Generationcited by 117
  10. 2026SWE-Skills-Bench: Do Agent Skills Actually Help in Real-World Software Engineering?cited by 72
  11. 2026XSkill: Continual Learning from Experience and Skills in Multimodal Agentscited by 71
  12. 2024OfficeBench: Benchmarking Language Agents across Multiple Applications for Office Automationcited by 66
9 more
  1. 2025Contextual Experience Replay for Self-Improvement of Language Agentscited by 46
  2. 2022VRDU: A Benchmark for Visually-rich Document Understandingcited by 35
  3. 2026When Single-Agent with Skills Replace Multi-Agent Systems and When They Failcited by 34
  4. 2024Investigate-Consolidate-Exploit: A General Strategy for Inter-Task Agent Self-Evolutioncited by 28
  5. 2024Skill Set Optimization: Reinforcing Language Model Behavior via Transferable Skillscited by 24
  6. 2024MIRAGE: Evaluating and Explaining Inductive Reasoning Process in Language Modelscited by 17
  7. 2025Talk to Your Slides: High-Efficiency Slide Editing via Language-Driven Structured Data Manipulationcited by 7
  8. 2026VAREX: A Benchmark for Multi-Modal Structured Extraction from Documentscited by 4
  9. 2025Co-DETECT: Collaborative Discovery of Edge Cases in Text Classificationcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 161.

161 papers.

Sort:

By area:

  1. 2026SkillOpt: Executive Strategy for Self-Evolving Agent Skillscited by 122
  2. 2026SkillX: Automatically Constructing Skill Knowledge Bases for Agentscited by 69
  3. 2026SkillLearnBench: Benchmarking Continual Learning Methods for Agent Skill Generation on Real-World Taskscited by 45
  4. 2026SkillGen: Verified Inference-Time Agent Skill Synthesiscited by 29
  5. 2026SkillGrad: Optimizing Agent Skills Like Gradient Descentcited by 23
  6. 2026From Failed Trajectories to Reliable LLM Agents: Diagnosing and Repairing Harness Flawscited by 21
  7. 2026SkillFlow:Benchmarking Lifelong Skill Discovery and Evolution for Autonomous Agentscited by 19
  8. 2026From Storage to Experience: A Survey on the Evolution of LLM Agent Memory Mechanismscited by 19
  9. 2026SkillsVote: Lifecycle Governance of Agent Skills from Collection, Recommendation to Evolutioncited by 18
  10. 2026From Raw Experience to Skill Consumption: A Systematic Study of Model-Generated Agent Skillscited by 18
  11. 2026SkillOpt-Lite: Better and Faster Agent Self-evolution via One Line of Vibecited by 18
  12. 2026A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applicationscited by 17
38 more
  1. 2026SkillGenBench: Benchmarking Skill Generation Pipelines for LLM Agentscited by 17
  2. 2026Experience Compression Spectrum: Unifying Memory, Skills, and Rules in LLM Agentscited by 16
  3. 2026ReSkill: Reconciling Skill Creation with Policy Optimization in Agentic RLcited by 13
  4. 2026SkillEvolver: Skill Learning as a Meta-Skillcited by 12
  5. 2026Evo-Harness: Context-to-Harness Skill Compilation for Self-Evolving Agentscited by 12
  6. 2026SkillAudit: Ground-Truth-Free Skill Evolution via Paired Trajectory Auditingcited by 11
  7. 2026Dynamic Skill Lifecycle Management for Agentic Reinforcement Learningcited by 10
  8. 2026Skills on the Fly: Test-Time Adaptive Skill Synthesis for LLM Agentscited by 10
  9. 2026SkillRAE: Agent Skill-Based Context Compilation for Retrieval-Augmented Executioncited by 8
  10. 2026EvoAgentBench: Benchmarking Agent Self-Evolution via Ability Transfercited by 8
  11. 2026WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolutioncited by 7
  12. 2026From Multi-Agent to Single-Agent: When Is Skill Distillation Beneficial?cited by 6
  13. 2026CODESKILL: Learning Self-Evolving Skills for Coding Agentscited by 6
  14. 2026Agent Skill Evaluation and Evolution: Frameworks and Benchmarkscited by 6
  15. 2026Dynamic Agent Skills: A Lifecycle Survey and Taxonomy of Evolving Skill Librariescited by 6
  16. 2026Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving Skillscited by 6
  17. 2026VideoWeaver: Evaluating and Evolving Skills for Agentic Long Video Generationcited by 5
  18. 2026Co-Evolving Skill Generation and Policy Optimizationcited by 5
  19. 2026SkillCAT: Contrastive, Assessment-Augmented and Topology-AwareSkill Self-Evolution for LLM Agentscited by 5
  20. 2026COMFYCLAW: Self-Evolving Skill Harnesses for Image Generation Workflowscited by 5
  21. 2026What Resolve Rate Hides: Trajectory Structure Diagnostics for Coding Agentscited by 5
  22. 2026DarwinX: Evolving Agent Harnesses Through Natural Selectioncited by 5
  23. 2026The Last AI Built by Humans: Toward Genuine Recursive Self-Improvementcited by 5
  24. 2026ClawTrace: Cost-Aware Tracing for LLM Agent Skill Distillationcited by 4
  25. 2026Rewarding the Scientific Process: Process-Level Reward Modeling for Agentic Data Analysiscited by 4
  26. 2026SkillSmith: Co-Evolving Skills and Tools for Self-Improving Agent Systemscited by 4
  27. 2026Agent Skills Matter: Inferring Proprietary Skills from Execution Trajectoriescited by 4
  28. 2026SkillCommit: Evolving Agent Skills through Behaviorally Validated Scope Expansioncited by 4
  29. 2026SkillMAS: Skill Co-Evolution with LLM-based Multi-Agent Systemcited by 3
  30. 2026ArtisanCAD: An Industrial-Level CAD Agent with Expert-Grounded Knowledge Distillationcited by 3
  31. 2026The Regression Tax: Decomposing Why Skills Help and Hurt LLM Agentscited by 3
  32. 2026SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolutioncited by 3
  33. 2026Rethinking Self-Evolving Agent Skills: Feedback Dynamics over Multiple Roundscited by 3
  34. 2026KV-Skill: Forging Expertise in the Model's Native Languagecited by 3
  35. 2026MIND-Skill: Quality-Guaranteed Skill Generation via Multi-Agent Induction and Deductioncited by 2
  36. 2026Ace-Skill: Bootstrapping Multimodal Agents with Prioritized and Clustered Evolutioncited by 2
  37. 2026OptSkills: Learning Generalizable Optimization Skills from Problem Archetypes via Cluster-Based Distillationcited by 2
  38. 2026SkillComposer: Learning to Evolve Agent Skills for Specification and Generalizationcited by 2

50 most cited of 161 All 161 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories
  2. 2026SKT: Skill-Use Training at Scale via Verified Synthetic Data Generation
  3. 2026SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving
  4. 2026MUSE-Autoskill: Self-Evolving Agents via Skill Creation, Memory, Management, and Evaluation
  5. 2026Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward
  6. 2026SkillGym: Training Skill-Use Agents with Automatic Verifiable Environment Generation
  7. 2026SkillJuror: Measuring How Agent Skill Organization Changes Runtime Behavior
  8. 2026HiSkill: Empowering LLM Agents with Hierarchical Skill Graphs