mlsanityData to 7 Oct 2026

Agent skill learning and evolution

259 papers10 this weekMedium confidence in this grouping: mostly one topic, with blurry edgesMedium confidence in this grouping: mostly one topic, with blurry edges · About

Keywords: skill, skills, agent, agents, task

Key facts. Agent skill learning and evolution: 259 papers on arXiv since 2024 (data through 7 Oct 2026). 87 of its 259 mapped papers (34%) cite “SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks” (2026, arXiv 2602.12670). 65 of its 259 mapped papers (25%) cite “SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning” (2026, arXiv 2602.08234). 60 new papers 17 Sep–7 Oct.

What’s new

17 Sep – 7 Oct 2026

60 new papers, 17 Sep – 7 Oct; 10 this week.

  1. This week · 10
  2. Skill-SLM: Agent Skill-driven Small Language Models for Reliable Robot Operation
  3. Before They Can Solve: Predicting Post-Training Coding-Agent Performance from Base Models also fits: Software engineering coding agents
  4. UniSkill: Learning Actor-Aligned Skill Proposals for an Evolving Policy
  5. Self-Evolve With a Reference:Anchored Training of Tool-Integrated Agents also fits: Reinforcement learning for LLM agents
  6. SkillForge: Co-Evolving Skills and Agents via Dynamic Skill Lifecycles
  7. Goldsmith: Gold-Loss-Guided Definition Optimization with an Agentic Annotation Harness also fits: Software engineering coding agents
  8. An Empirical Study of Agent Skills' Downstream Utility
  9. Learn2Play Bench: How Well Do LLM Agents Learn from Experience in Unfamiliar Environments? also fits: Reinforcement learning for LLM agents
  10. DAEDALUS: Bootstrapping Agent Memory from Self-Generated Tasks also fits: Software engineering coding agents
  11. SkillPoison: Progressive Skill Poisoning via Successful Experiences
  12. Earlier · 50
All 60
  1. TeleTune: Evolving Agent Skills From Offline Telemetry
  2. ASCENT: Online Test-Time Training of Long-Horizon Agents via Self-Distillation of Verified Experience also fits: Reinforcement learning for LLM agents
  3. EVISKILL: Grounding Skill Evolution in Replayable Evidence
  4. Self-Evaluating Recursive Agents also fits: General LLM agent frameworks and benchmarks
  5. SkillScriptBench: Benchmarking Self-Evolution of Executable Agent Skill Packages Beyond Markdown also fits: Software engineering coding agents
  6. Sentry: Learning to Recover from LLM Agent Failures at Test Time
  7. Skill2Real: Agentic Skill Learning for Zero-Shot Sim-to-Real Robot Manipulation
  8. Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents
  9. Continuous Process-Level Evaluation for Evolving Enterprise AI Agent Skills also fits: Software engineering coding agents
  10. SkillEvoLean: Mutation-enhanced skill evolution for Lean provers
  11. Chaining Skills to Hijack LLM Agents
  12. MCRI: A Four-Dimensional Framework for Analyzing and Evaluating Agent Skills
  13. What Does a Skill Actually Do? Estimands and Evaluation Validity for Tool and Skill Use in LLM Agents: A Critical Review
  14. SkillSpec: Consensus-Gated Agent Skill Evolution via Representation Specialization
  15. Pretext: Defeating Malicious Skill Detection Frameworks for AI Agents also fits: Agent threats, guardrails and governance
  16. SkillFM: Generating Skills for LLM Agents via Latent Flow Matching
  17. WorkGenesis: Building the Worlds That Teach Agents to Work
  18. ReSAIL: Mitigating Collapse in Iterative Agent Self-Distillation also fits: Reinforcement learning for LLM agents
  19. Rep2Skill: Representation-Guided Skill Self-Evolution for LLM Agents
  20. SkillSeek: Revisiting Agent Skill Retrieval at Marketplace Scale
  21. SelfSearch: Reward-Free Search for Self-Improving Agents
  22. SkillGym: Training Skill-Use Agents with Automatic Verifiable Environment Generation
  23. WEFT: Scaling Tool-Use Post-Training for General-Purpose Agents also fits: Reinforcement learning for LLM agents
  24. Distilling Agentic Systems: A Roadmap across Models, Artifacts, and Harnesses also fits: General LLM agent frameworks and benchmarks
  25. Semantic Projection for Continual Self-Evolution of Language Agents also fits: Reinforcement learning for LLM agents
  26. SEABench: Benchmarking Endogenous Misalignment In Self-Evolving Agents also fits: General LLM agent frameworks and benchmarks
  27. SkillWeaver: Agentic Exploration over Neural Interaction Skills for Scalable Robot Data Generation
  28. SAGE: A Statistical Acceptance Gate for Self-Evolving Agents also fits: General LLM agent frameworks and benchmarks
  29. Beyond Skill Evolution: Self-Evolving Context Management Policies for Long-Horizon Agent Harnesses also fits: Reinforcement learning for LLM agents
  30. SkillFocus: Evolving Agent Skills via Capability Decomposition
  31. Maintaining Benchmarks Against Increasingly Capable Agents: Detection and Remediation of Unearned Passes also fits: Software engineering coding agents
  32. When Successful Strategies Fail: Adaptation to Environmental Novelty in Terminal Agents also fits: Reinforcement learning for LLM agents
  33. R2 Flow: Recursive Self-Improvement via Recursive Skill Evolution
  34. X-Tree: Tokenizing Reusable Experience for Efficient Agent Generalization
  35. Certified Long-Horizon Code Agent Evolution via Validation-Gated Skill Optimization also fits: Software engineering coding agents
  36. RoboFoundry: System-as-Policy Evolution for Self-Learning Embodied Agents also fits: Reinforcement learning for LLM agents
  37. SkillVine: Agent Skill Evolution via Branching Exploration
  38. ExpVoyager: Direct Experience Navigation for Dynamic Agent Skill Synthesis
  39. Enabling Timely Guidance before Skill Retrieval: Retaining Helpful Warm Tips in Agent Context
  40. Up and Down the Abstraction Ladder: Code-Based Skills for Language Agents also fits: Reinforcement learning for LLM agents
  41. SkillEvoReg: Regularizing Agent Skill Evolution Against Overfitting
  42. A Wrong Turn Does Not Ruin the Journey: Deviation-Guided Skill Self-Evolution for LLM Agents
  43. Scope Before You Persist: Preventing Cross-Family Interference in Agent Memory also fits: Agent memory systems
  44. Progressive Skill Discovery as Access Control for Tool-Using LLM Agents: Structural Governance through Role-Scoped Capability Delivery also fits: Tool learning and function calling
  45. ChipMEM: Verification-Grounded Memory for EDA Agents
  46. SkillApt: Learning When to Activate Agent Skills from Counterfactual Evidence
  47. RRSI: Regularized Recursive Self-Improvement of Agent Harnesses also fits: Reinforcement learning for LLM agents
  48. From Lexical Baselines to Agentic Retrieval-Augmented Generation: Structured Skill and Responsibility-Level Extraction with the SFIA Framework
  49. GraphSkillEvo: Evolutionary Optimization of Graph-Structured Agent Skills
  50. EconSkills: Studying Skill Transfer and Retrieval for Web Agents on Live Economic Data

Foundations

Most cited by papers here

  1. SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks2026cited by 87 here
  2. SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning2026cited by 65 here
  3. Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills2026cited by 46 here
  4. Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward2026cited by 38 here
  5. SoK: Agentic Skills -- Beyond Tool Use in LLM Agents2026cited by 38 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 97 here
  2. 2023WebArena: A Realistic Web Environment for Building Autonomous Agentscited by 35 here
  3. 2024Agent Workflow Memorycited by 28 here
  4. 2025SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skillscited by 33 here
  5. 2025Reinforcement Learning for Self-Improving Agent with Skill Librarycited by 37 here
  6. 2026Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forwardcited by 38 here
  7. 2026Organizing, Orchestrating, and Benchmarking Agent Skills at Ecosystem Scalecited by 32 here
  8. 2026Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skillscited by 46 here
16 more it builds on, not in this chain
  1. 2022WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agentscited by 34 here
  2. 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 78 here
  3. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 86 here
  4. 2023Gorilla: Large Language Model Connected with Massive APIscited by 38 here
  5. 2023ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIscited by 52 here
  6. 2023ExpeL: LLM Agents Are Experiential Learnerscited by 75 here
  7. 2024SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineeringcited by 33 here
  8. 2026MemSkill: Learning and Evolving Memory Skills for Self-Evolving Agentscited by 28 here
  9. 2026SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learningcited by 65 here
  10. 2026SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Taskscited by 87 here
  11. 2026SoK: Agentic Skills -- Beyond Tool Use in LLM Agentscited by 38 here
  12. 2026AutoSkill: Experience-Driven Lifelong Learning via Skill Self-Evolutioncited by 37 here
  13. 2026EvoSkill: Automated Skill Discovery for Multi-Agent Systemscited by 34 here
  14. 2026CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verificationcited by 38 here
  15. 2026How Well Do Agentic Skills Work in the Wild: Benchmarking LLM Skill Usage in Realistic Settingscited by 35 here
  16. 2026Graph-of-Skills: Dependency-Aware Structural Retrieval for Massive Agent Skillscited by 30 here

The 24 most cited of 907.

Browse

30 of 259, ranked 6–35

  1. 6CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification202638
  2. 7Reinforcement Learning for Self-Improving Agent with Skill Library202537
  3. 8How Well Do Agentic Skills Work in the Wild: Benchmarking LLM Skill Usage in Realistic Settings202635
  4. 9EvoSkill: Automated Skill Discovery for Multi-Agent Systems202634
  5. 10Organizing, Orchestrating, and Benchmarking Agent Skills at Ecosystem Scale202632
  6. 11Graph-of-Skills: Dependency-Aware Structural Retrieval for Massive Agent Skills202630
  7. 12SkillRouter: Skill Routing for LLM Agents at Scale202628
  8. 13SkillX: Automatically Constructing Skill Knowledge Bases for Agents202628
  9. 14SkillClaw: Let Skills Evolve Collectively with Agentic Evolver202624
  10. 15SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization202622
All 30
  1. 16Agent Skills: A Data-Driven Analysis of Claude Skills for Extending Large Language Model Functionality202620
  2. 17SkillRet: A Large-Scale Benchmark for Skill Retrieval in LLM Agents202618
  3. 18SkillOpt: Executive Strategy for Self-Evolving Agent Skills202618
  4. 19SkillLearnBench: Benchmarking Continual Learning Methods for Agent Skill Generation on Real-World Tasks202617
  5. 20Skill Retrieval Augmentation for Agentic AI202616
  6. 21GraSP: Graph-Structured Skill Compositions for LLM Agents202615
  7. 22SkillOS: Learning Skill Curation for Self-Evolving Agents202615
  8. 23Skill-Pro: Learning Reusable Skills from Experience via Non-Parametric PPO for LLM Agents202614
  9. 24SkillReducer: Optimizing LLM Agent Skills for Token Efficiency202613
  10. 25SkillCraft: Can LLM Agents Learn to Use Tools Skillfully?202612
  11. 26ARISE: Agent Reasoning with Intrinsic Skill Evolution in Hierarchical Reinforcement Learning202611
  12. 27Dynamic Dual-Granularity Skill Bank for Agentic RL202611
  13. 28SkillForge: Forging Domain-Specific, Self-Evolving Agent Skills in Cloud Technical Support202610
  14. 29SkillFlow: Scalable and Efficient Agent Skill Retrieval System202510
  15. 30MetaClaw: Just Talk -- An Agent That Meta-Learns and Evolves in the Wild20269
  16. 31SkVM: Revisiting Language VM for Skills across Heterogenous LLMs and Harnesses20268
  17. 32SkillGraph: Skill-Augmented Reinforcement Learning for Agents via Evolving Skill Graphs20268
  18. 33Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning20267
  19. 34A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applications20267
  20. 35SkillGen: Verified Inference-Time Agent Skill Synthesis20267