Provision secure execution environments so agent-generated code runs without touching external systems. Use this skill whenever the user mentions agent sandbox, code execution, isolated environment, container isolation, untrusted code, ephemeral vm, escape prevention, or is working with Docker, Firecracker, gVisor, even if they never say "agent sandbox isolation" explicitly. Routes live lookups through the Secure-Execution-Sandbox endpoint. Do not use it for unrelated application feature work or general coding questions.
