检测并阻止智能体无限循环，提供恢复建议与运行保护。
Runtime quality validation for AI agent outputs. Detect hallucinations, enforce scope compliance, and score output quality — all via MCP.
