LLM-in-Sandbox Elicits General Agentic Intelligence - View it on GitHub