vercel-ai-sdk
Pass
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [Subprocess Execution for MCP Tools]: The skill demonstrates using
ai.mcp.get_stdio_toolsto run Model Context Protocol (MCP) servers using local commands likenpx. This enables the agent to interact with the local environment, such as the filesystem, which is a core feature of the protocol. - [Remote Tool Configuration]: Through
ai.mcp.get_http_tools, the skill can retrieve tool schemas and endpoint information from external URLs. This allows for dynamic integration with remote services and APIs. - [Indirect Prompt Injection Surface]: As an agentic framework, the system is designed to ingest and act upon data from external sources (LLMs and tools). This creates a potential surface for indirect prompt injection, an inherent characteristic of applications that process untrusted data through an LLM. Developers should follow best practices for sanitizing tool outputs.
Audit Metadata