Create an automated sandboxing layer for LLM agents that restricts file system access and network egress at the kernel level. This ensures that agentic workflows can run third-party code without risking the host environment.
Suggested repo: sandboxAgent
"Let your AI agents execute arbitrary code without burning down your server."
Estimated effort: 80h