Saved in:
Bibliographic Details
Main Authors: Zhou, Yifan, Zhang, Zhentao, Cheng, Ziming, Zhang, Shuo, Lan, Qizhen, Chen, Zhangquan, Yang, Zhi, QianyuXu, Chen, Ronghao, Wang, Huacan, Hu, Sen
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.18693
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910232983109632
author Zhou, Yifan
Zhang, Zhentao
Cheng, Ziming
Zhang, Shuo
Lan, Qizhen
Chen, Zhangquan
Yang, Zhi
QianyuXu
Chen, Ronghao
Wang, Huacan
Hu, Sen
author_facet Zhou, Yifan
Zhang, Zhentao
Cheng, Ziming
Zhang, Shuo
Lan, Qizhen
Chen, Zhangquan
Yang, Zhi
QianyuXu
Chen, Ronghao
Wang, Huacan
Hu, Sen
contents As LLM agents are increasingly built around reusable skills, a central challenge is no longer only whether agents can use provided skills, but whether they can generate correct, reusable, and executable skills from repositories and documents. Existing benchmarks primarily evaluate the efficacy of given skills or the ability of agents to solve downstream tasks from raw context, but they do not isolate skill generation itself as the object of study. We introduce SkillGenBench, a benchmark for evaluating skill generation pipelines under a unified and controlled protocol. In SkillGenBench, a generator receives raw corpora and produces standardized skill artifacts, which are then executed under fixed harnesses and assessed with unified evaluation procedures. The benchmark covers two generation regimes: task-conditioned generation, where a task-specific skill is synthesized after the task is revealed, and task-agnostic generation, where a reusable skill library must be distilled before downstream tasks are known. It also spans two complementary procedural sources: repository-grounded instances, where procedures are distributed across code, configuration, and scripts, and document-grounded instances, where procedures and constraints must be distilled from long-form text. We provide standardized task specifications, pinned environments, and evaluation protocols centered on deterministic execution-based checks, supplemented by auxiliary signals for diagnosis. Experiments across a range of skill-generation methods and backbones show substantial performance variation, highlight the difficulty of reusable skill distillation, and reveal distinct failure modes in skill generation from software repositories versus long-form documents. SkillGenBench establishes a reproducible testbed for studying skill generation as an independent research problem in agent systems.
format Preprint
id arxiv_https___arxiv_org_abs_2605_18693
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SkillGenBench: Benchmarking Skill Generation Pipelines for LLM Agents
Zhou, Yifan
Zhang, Zhentao
Cheng, Ziming
Zhang, Shuo
Lan, Qizhen
Chen, Zhangquan
Yang, Zhi
QianyuXu
Chen, Ronghao
Wang, Huacan
Hu, Sen
Artificial Intelligence
As LLM agents are increasingly built around reusable skills, a central challenge is no longer only whether agents can use provided skills, but whether they can generate correct, reusable, and executable skills from repositories and documents. Existing benchmarks primarily evaluate the efficacy of given skills or the ability of agents to solve downstream tasks from raw context, but they do not isolate skill generation itself as the object of study. We introduce SkillGenBench, a benchmark for evaluating skill generation pipelines under a unified and controlled protocol. In SkillGenBench, a generator receives raw corpora and produces standardized skill artifacts, which are then executed under fixed harnesses and assessed with unified evaluation procedures. The benchmark covers two generation regimes: task-conditioned generation, where a task-specific skill is synthesized after the task is revealed, and task-agnostic generation, where a reusable skill library must be distilled before downstream tasks are known. It also spans two complementary procedural sources: repository-grounded instances, where procedures are distributed across code, configuration, and scripts, and document-grounded instances, where procedures and constraints must be distilled from long-form text. We provide standardized task specifications, pinned environments, and evaluation protocols centered on deterministic execution-based checks, supplemented by auxiliary signals for diagnosis. Experiments across a range of skill-generation methods and backbones show substantial performance variation, highlight the difficulty of reusable skill distillation, and reveal distinct failure modes in skill generation from software repositories versus long-form documents. SkillGenBench establishes a reproducible testbed for studying skill generation as an independent research problem in agent systems.
title SkillGenBench: Benchmarking Skill Generation Pipelines for LLM Agents
topic Artificial Intelligence
url https://arxiv.org/abs/2605.18693