"author:ShinYeol-Lee" matching MCP servers:

Provides a human-in-the-loop security layer for AI agents by intercepting file operations, explaining them with a local LLM, and enforcing a deterministic policy that requires user approval for risky actions.