# Organizing, Orchestrating, and Benchmarking Agent Skills at Ecosystem Scale

Source: [Organizing, Orchestrating, and Benchmarking Agent Skills at Ecosystem Scale](<https://www.sota2.com/research/paper/69a66c582d603e6de3e3d194>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-04T05:00:19.274Z

Authors: Hao Li, Chunjiang Mu, Jianhao Chen, Siyue Ren, Zhiyao Cui, Yiqun Zhang, Lei Bai, Shuyue Hu

Published: 2026-03-02T18:46:47.000Z

## Abstract

The rapid proliferation of Claude agent skills has raised the central question of how to effectively leverage, manage, and scale the agent skill ecosystem. In this paper, we propose AgentSkillOS, the first principled framework for skill selection, orchestration, and ecosystem-level management. AgentSkillOS comprises two stages: (i) Manage Skills, which organizes skills into a capability tree via node-level recursive categorization for efficient discovery; and (ii) Solve Tasks, which retrieves, orchestrates, and executes multiple skills through DAG-based pipelines. To evaluate the agent's ability to invoke skills, we construct a benchmark of 30 artifact-rich tasks across five categories: data computation, document creation, motion video, visual design, and web interaction. We assess the quality of task outputs using LLM-based pairwise evaluation, and the results are aggregated via a Bradley-Terry model to produce unified quality scores. Experiments across three skill ecosystem scales (200 to 200K skills) show that tree-based retrieval effectively approximates oracle skill selection, and that DAG-based orchestration substantially outperforms native flat invocation even when given the identical skill set. Our findings confirm that structured composition is the key to unlocking skill potential. Our GitHub repository is available at:https://github.com/ynulihao/AgentSkillOS.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Skill invocation in LLM agents on 30 artifact-rich tasks (test)](<https://www.sota2.com/research/sota/skill-invocation-in-llm-agents-on-30-artifact-rich-tasks-test.md>): Skill invocation in LLM agents; 30 artifact-rich tasks (test); Data Computation Wins: 18.

- [Downstream task execution on AgentSkillOS](<https://www.sota2.com/research/sota/downstream-task-execution-on-agentskillos.md>): Downstream task execution; AgentSkillOS; Score: 83.5.

- [Downstream task execution on SkillsBench](<https://www.sota2.com/research/sota/downstream-task-execution-on-skillsbench.md>): Downstream task execution; SkillsBench; Reward Mean (%): 17.24.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.02176>)

[PDF](<https://arxiv.org/pdf/2603.02176.pdf>)
