Agent Skills is an emerging open standard that defines a modular, filesystem-based packaging format, allowing LLM-based agents to acquire domain-specific expertise on demand. The format has seen rapid adoption across multiple agentic systems, according to a new paper on arXiv.

The paper, titled "Towards Secure Agent Skills: Architecture, Threat Taxonomy, and Security Analysis," addresses the security implications of this standard. It presents an architecture for understanding how Agent Skills are structured and deployed, and introduces a threat taxonomy to categorize potential vulnerabilities.

Because the abstract is brief, the paper's specific findings are not detailed here. The authors emphasize the need for systematic security analysis as the standard gains traction, suggesting that the proposed taxonomy could serve as a foundation for securing future Agent Skills implementations.