SkillFlow:Benchmarking Lifelong Skill Discovery and Evolution for Autonomous Agents

• URL: http://arxiv.org/abs/2604.17308v1
• Date: Sun, 19 Apr 2026 07:51:46 GMT
• Title: SkillFlow:Benchmarking Lifelong Skill Discovery and Evolution for Autonomous Agents
• Abstract summary: We introduce SkillFlow, a benchmark of 166 tasks across 20 families in which task construction within each family follows a Domain-Agnostic Execution Flow (DAEF)<n>Agents are evaluated under an Agentic Lifelong Learning protocol in which they begin without skills, solve tasks sequentially within each family, externalize lessons through trajectory- and rubric-driven skill patches, and carry the updated library forward.
• Score: 33.117459452689964
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: As the capability frontier of autonomous agents continues to expand, they are increasingly able to complete specialized tasks through plug-and-play external skills. Yet current benchmarks mostly test whether models can use provided skills, leaving open whether they can discover skills from experience, repair them after failure, and maintain a coherent library over time. We introduce SkillFlow, a benchmark of 166 tasks across 20 families in which task construction within each family follows a Domain-Agnostic Execution Flow (DAEF) that defines an agent workflow framework, allowing these tasks to share a consistent workflow. Agents are evaluated under an Agentic Lifelong Learning protocol in which they begin without skills, solve tasks sequentially within each family, externalize lessons through trajectory- and rubric-driven skill patches, and carry the updated library forward. Experiments reveal a substantial capability gap. For Claude Opus 4.6, lifelong skill evolution improves task success from 62.65% to 71.08% (+8.43 points). However, high skill usage does not necessarily imply high utility: Kimi K2.5 gains only +0.60 points despite 66.87% skill usage, while Qwen-Coder-Next reaches only a 44.58% task completion rate and still regresses relative to the vanilla setting. SkillFlow contributes a structured testbed for this direction and an in-depth empirical analysis of skill discovery, patching, transfer, and their failure modes under lifelong evaluation.

Related papers

• Behavior-Skill: A Fine-Grained Benchmark for Evaluating Vision-Language-Action Policies in Long-Horizon Tasks [39.959375504916494]
  We present Behavior-Skill, a benchmark that reformulates the learning and evaluation of long-horizon tasks.<n>It contains 235,492 skill instances from 10,000 demonstrations across 50 household tasks and 34 semantic skill categories.<n>We show that Behavior-Skill complements full-task evaluation by exposing intermediate capability profiles for analyzing and improving long-horizon VLA policies.
  arXiv  Detail & Related papers  (2026-08-31T10:03:48Z)
• Evaluating Skills, Not Just Agents: Agentic Continuous Evaluation of Skills [9.077159743623076]
  repository-native framework for evaluating skills and product capability packages as executable agent artifacts.<n>ACES runs paired live trials with and without a target skill, grades six default runtime metrics, and reports Skill Lift.<n>Open-source implementation of the methodology is available in NVIDIA SkillEvaluator.
  arXiv  Detail & Related papers  (2026-08-20T23:26:14Z)
• Write, Execute, Refine: From Skill Followers to Skill Optimizers via Reinforcement Learning from Execution Feedback [52.78692582507693]
  Expert-written natural language skills can improve tool-using agents, yet agent-authored skills perform 8-11 points worse than using no skill.<n>We study how to organize execution experience from intermediate skills into training states for an executor.
  arXiv  Detail & Related papers  (2026-08-18T09:52:48Z)
• Demystifying Agent Skills: Why They Work-Until They Don't [58.92875630998418]
  Existing evaluations largely measure whether skills improve aggregated task success, leaving a more fundamental question underexplored: emphtextbfWhen do skills help, why do they work, and where do they fail?<n>We isolate the effects of representation, outcome annotation, retrieval difficulty, and cross-framework of skills.<n>We consolidate these observations into a taxonomy of three high-level categories and twelve skill-use modes.
  arXiv  Detail & Related papers  (2026-08-14T07:26:38Z)
• SkillCoach: Self-Evolving Rubrics for Evaluating and Enhancing Agentic Skill-Use [21.05387636058484]
  SkillCoach is a self-evolving framework for evaluating and enhancing agentic skill-use.<n>It derives skill-grounded process rubrics from real rollouts and evaluates trajectories along four dimensions.<n>It keeps the external verifier as a separate outcome signal, allowing process quality to be distinguished from accidental task success.
  arXiv  Detail & Related papers  (2026-07-02T08:28:51Z)
• SkillAudit: Ground-Truth-Free Skill Evolution via Paired Trajectory Auditing [81.51044612408793]
  SkillAudit is a framework for evolving agent skills without ground-truth feedback.<n>To turn behavioral differences into edit guidance, SkillAudit uses Process-Aligned Contrastive Evaluation.<n>It routes edits through two pipelines: Refine removes noisy or irrelevant guidance from broadly useful skills, while Repair replaces passages that conflict with the task.
  arXiv  Detail & Related papers  (2026-06-12T08:20:09Z)
• SkillOS: Learning Skill Curation for Self-Evolving Agents [67.94374107466957]
  We propose SkillOS, an experience-driven RL training recipe for learning skill curation in self-evolving agents.<n>SkillOS pairs a frozen agent executor that retrieves and applies skills with a trainable skill curator that updates an external SkillRepo from accumulated experience.<n>We show that SkillOS consistently outperforms memory-free and strong memory-based baselines in both effectiveness and efficiency.
  arXiv  Detail & Related papers  (2026-05-07T17:31:50Z)
• AcademiClaw: When Students Set Challenges for AI Agents [69.70303995121154]
  AcademiClaw is a benchmark of 80 complex, long-horizon tasks sourced directly from university students' real academic disconnects.<n>We show that even the best of models achieves only a 55% pass rate.<n>We hope that AcademiClaw and its open-sourced data and code can serve as a useful resource for the OpenClaw community.
  arXiv  Detail & Related papers  (2026-05-04T14:40:42Z)
• How Well Do Agentic Skills Work in the Wild: Benchmarking LLM Skill Usage in Realistic Settings [29.3744517703302]
  We study the first comprehensive study of skill utility under progressively challenging realistic settings.<n>Our findings reveal that performance gains degrade consistently as settings become more realistic.<n>We show that query-specific refinement substantially recovers lost performance when the initial skills are of reasonable relevance and quality.
  arXiv  Detail & Related papers  (2026-04-06T00:10:30Z)
• SKILLFOUNDRY: Building Self-Evolving Agent Skill Libraries from Heterogeneous Scientific Resources [5.7282243774488295]
  We present SkillFoundry, a self-evolving framework that converts resources into validated agent skills.<n>We show that SkillFoundry produces a substantially novel and internally valid skill library.<n>We also show that SkillFoundry can design new task-specific skills on demand for concrete scientific objectives.
  arXiv  Detail & Related papers  (2026-04-05T05:02:18Z)
• SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization [61.37478254738943]
  Inference-time skill augmentation is fundamentally limited.<n>We introduce SKILL0, an in-context reinforcement learning framework designed for skill internalization.<n>SKILL0 achieves substantial improvements over the standard RL baseline.
  arXiv  Detail & Related papers  (2026-04-02T17:03:05Z)
• EvoSkill: Automated Skill Discovery for Multi-Agent Systems [6.319876096746374]
  We introduce textbfEvoSkill, a self-evolving framework that automatically discovers and refines agent skills.<n>EvoSkill analyzes execution failures, proposes new skills or edits to existing ones, and materializes them into structured, reusable skill folders.<n>We evaluate EvoSkill on two benchmarks: OfficeQA, a grounded reasoning benchmark over U.S. Treasury data, and SealQA, a noisy retrieval benchmark.
  arXiv  Detail & Related papers  (2026-03-03T09:07:22Z)
• SkillCraft: Can LLM Agents Learn to Use Tools Skillfully? [67.69996753743129]
  We introduce SkillCraft, a benchmark explicitly stress-test agent ability to form and reuse higher-level tool compositions.<n> SkillCraft features realistic, highly compositional tool-use scenarios with difficulty scaled along both quantitative and structural dimensions.<n>We propose a lightweight evaluation protocol that enables agents to auto-compose atomic tools into executable Skills, cache and reuse them inside and across tasks.
  arXiv  Detail & Related papers  (2026-02-28T15:44:31Z)
• SkillNet: Create, Evaluate, and Connect AI Skills [159.47504178122156]
  SkillNet is an open infrastructure designed to create, evaluate, and organize AI skills at scale.<n>Our infrastructure integrates a repository of over 200,000 skills, an interactive platform, and a versatile Python toolkit.
  arXiv  Detail & Related papers  (2026-02-26T14:24:02Z)
• SoK: Agentic Skills -- Beyond Tool Use in LLM Agents [6.356997609995175]
  Agentic systems increasingly rely on reusable procedural capabilities, textita.k.a., agentic skills, to execute long-horizon reliably.<n>This paper maps the skill layer across the full lifecycle (discovery, practice, distillation, storage, composition, evaluation, and update)<n>We analyze the security and governance implications of skill-based agents, covering supply-chain risks, prompt injection via skill payloads, and trust-tiered execution.
  arXiv  Detail & Related papers  (2026-02-24T13:11:38Z)
• SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks [61.89812116484928]
  Agent Skills are structured packages of procedural knowledge that augment LLM agents at inference time.<n>We present SkillsBench, a benchmark of 86 tasks across 11 domains paired with curated Skills and deterministic verifiers.<n>We show curated Skills raise average pass rate by 16.2 percentage points(pp), but effects vary widely by domain.<n>Self-generated Skills provide no benefit on average, showing that models cannot reliably author the procedural knowledge they benefit from consuming.
  arXiv  Detail & Related papers  (2026-02-13T07:06:06Z)
• Hierarchical Skills for Efficient Exploration [70.62309286348057]
  In reinforcement learning, pre-trained low-level skills have the potential to greatly facilitate exploration. Prior knowledge of the downstream task is required to strike the right balance between generality (fine-grained control) and specificity (faster learning) in skill design. We propose a hierarchical skill learning framework that acquires skills of varying complexity in an unsupervised manner.
  arXiv  Detail & Related papers  (2021-10-20T22:29:32Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.