Audit Model Context Protocol (MCP) server definitions by classifying filesystem paths, checking network endpoints, flagging shell execution privileges, and calculating blast radius scores before granting client access.
MCP servers run locally or in cloud containers with direct access to user tools, credentials, filesystems, and CLI commands. Unaudited MCP server configs often request wildcards (`*`) or unrestricted directory mounts, allowing LLM agent tool calls to read sensitive `.env` files or execute unverified shell scripts.
Audit this pattern live with PromptImageLab developer tools:
Explore Tools