Analysis of the Attack Surface in the Agent Skills Architecture: Case Studies and Ecosystem Research
fevereiro 3, 2026
Background As LLMs and intelligent agents expand from dialogue to task execution, the encapsulation, reuse and orchestration of LLM capabilities have become key issues. As a capability abstraction mechanism, Skills encapsulates reasoning logic, tool calls and execution processes into reusable skill units, enabling the model to achieve stable, consistent and manageable operations when performing complex […]
