Ask HN: AI Agent and harness containerization/security recommendations

A user is seeking advice on safely using AI tools that require access to local project and user folders, as well as operating system files. The goal is to protect system and data integrity while maintaining ease of use. The user is looking for recommendations on containerization and security strategies for AI agents. They are currently using OpenCode and DeepSeek on Linux Mint Debian Edition and are open to suggestions on operating system, agent harness, and security tool stacks. The user proposes organizing comments by these categories to facilitate discussion. They aim to find a balance between security and productivity in their workflow.