GitHub - mgasparin67-cmd/Agent-verification_Assure：面向 AI 智能体的独立保障平面：策略执行网关、确定性断言、模型基准测试。不采用 LLM 作为裁判。

GitHub - mgasparin67-cmd/Agent-verification_Assure: Independent assurance plane for AI agents: policy-enforcing gateway， deterministic assertions， model benchmarks. No LLM-as-judge.

Assure is a verification framework that runs AI agents against a disposable sandbox, enforcing policy through a gateway and validating results with deterministic assertions, without relying on LLM-based judgment.