![[2602.12430] Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.12430] Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward
Summary
This paper discusses the evolution of large language models (LLMs) into modular agents equipped with skills, emphasizing architecture, acquisition, security, and future directions.
Why It Matters
As LLMs transition from monolithic structures to modular agents, understanding the implications of agent skills is crucial for advancing AI capabilities. This paper outlines the architectural foundations and security challenges, providing a roadmap for future research and development in AI systems.
Key Takeaways
- Agent skills allow for dynamic capability extension of LLMs without retraining.
- The paper identifies significant security vulnerabilities in community-contributed skills.
- A proposed Skill Trust and Lifecycle Governance Framework aims to enhance skill deployment security.
- Seven open challenges are outlined for future research in skill ecosystems.
- The work focuses specifically on the skill abstraction layer, differentiating it from broader LLM surveys.
Computer Science > Multiagent Systems arXiv:2602.12430 (cs) [Submitted on 12 Feb 2026] Title:Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward Authors:Renjun Xu, Yang Yan View a PDF of the paper titled Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward, by Renjun Xu and Yang Yan View PDF HTML (experimental) Abstract:The transition from monolithic language models to modular, skill-equipped agents marks a defining shift in how large language models (LLMs) are deployed in practice. Rather than encoding all procedural knowledge within model weights, agent skills -- composable packages of instructions, code, and resources that agents load on demand -- enable dynamic capability extension without retraining. It is formalized in a paradigm of progressive disclosure, portable skill definitions, and integration with the Model Context Protocol (MCP). This survey provides a comprehensive treatment of the agent skills landscape, as it has rapidly evolved during the last few months. We organize the field along four axes: (i) architectural foundations, examining the this http URL specification, progressive context loading, and the complementary roles of skills and MCP; (ii) skill acquisition, covering reinforcement learning with skill libraries (SAGE), autonomous skill discovery (SEAgent), and compositional skill synthesis; (iii) deployment at scale, including the computer-use agent (CUA) stack...