What Happens When You Lock Claude in a macOS Sandbox

As AI coding assistants and CLI tools gain broader execution autonomy, constraining them within operating system boundaries becomes a practical necessity. This exploration examines running Claude inside a native macOS sandbox environment, tracing how it behaves when file system access, network requests, and system calls are strictly filtered. The experiment highlights the friction between agent capability and containment, offering a grounded look at securing agentic terminal workflows before unconstrained automation takes over developer machines.