I built an open-source zero-trust security runtime for AI agents. Here’s what I learned.

The problem nobody is talking about You give an AI agent access to your terminal. It can run shell...