Towards Secure Agent Skills: Architecture, Threat Taxonomy, and Security Analysis

Researchers conducted the first comprehensive security analysis of Agent Skills, an emerging standard for LLM-based agents to acquire domain expertise. The study identified significant structural vulnerabilities across the framework's lifecycle, including lack of data-instruction boundaries and insufficient security review processes.

• →Agent Skills framework has fundamental security flaws that cannot be fixed through incremental updates alone.
• →Seven threat categories and seventeen attack scenarios were identified across the Agent Skills ecosystem.
• →Five confirmed security incidents validate the identified threat taxonomy in real-world deployments.
• →The framework lacks mandatory marketplace security review and proper trust boundaries.
• →Most severe threats stem from structural properties including single-approval persistent trust models.