Running Claude Code dangerously (safely)
Summary
The article analyzes running Claude Code with a dangerous permissions flag and the safety implications. It compares Docker-based isolation with a VM approach using Vagrant and VirtualBox, offering a reproducible setup and a threat model. The piece provides practical steps and a reusable Vagrantfile to sandbox AI-driven automation safely.