TencentCloud/CubeSandbox ? reverse-engineered prompt

I want to build a high-performance sandbox service for running code generated by AI agents. Speed and security are the most important things. It needs to be extremely fast, with sandboxes starting up in milliseconds, and have a very small memory footprint—under 5MB per instance—so I can run thousands of them concurrently on a single server for high density.

For security, I don't want to just use Docker. It needs to have true hardware-level isolation using something like KVM to prevent container escapes and safely execute any code the LLM generates. The whole thing should be built in Rust for performance and safety.

A key feature is that it must be a drop-in replacement for the E2B service. That means it should be compatible with the E2B SDK, so users can switch over just by changing a URL. Make sure it can be deployed easily on a single node or scaled to a cluster. Please also include a quick-start guide and some examples.