Always sandbox your agent. - It prevents your agent from doing too much damage should an exploit exist. - The agent's built-in "sandboxing" causes agents to k

Always sandbox your agent. - It prevents your agent from doing too much damage should an exploit exist. - The agent's built-in "sandboxing" causes agents to keep asking permission for every damn thing, to the point where you just automatically answer "yes" to everything, and thus lose whatever benefits its sandbox had. It's why I wrote yoloAI: https://github.com/kstenerud/yoloai https://github.com/kstenerud/yoloai