How to Actually Run Your Coding Agent Safely (And Avoid the Horror Stories)

YouTube · Watch

• Addresses coding-agent security via Docker Sandboxes for isolation, directly responding to the home-directory-deletion/production-incident horror stories this project tracks in ai-agent-accountability.
• Demonstrates concrete risk of unrestricted agent filesystem access.
• Proposes microVM containerization providing “four isolation layers” as a practical safe-execution pattern without giving up agent capability.