Agent skill · software engineering · aradotso
cubesandbox-ai-sandbox
CubeSandbox — instant, hardware-isolated, E2B-compatible sandbox service for AI agents built on RustVMM/KVM
Why this skill is useful
Adds specific commands and workflows for managing isolated sandboxes using CubeSandbox that are not widely known.
What it needs
Requires curl, git installed locally. About 6k tokens when loaded. Last updated 2026-07-09. 70 stars on the source repository.
What this skill does
CubeSandbox AI Sandbox Skill Skill by ara.so — Daily 2026 Skills collection. CubeSandbox is a high-performance secure sandbox service built on RustVMM and KVM. It provides hardware-isolated (dedicated Guest OS kernel) sandbox environments that start in under 60ms, consume less than 5MB memory overhead per instance, and are fully compatible with the E2B SDK — making it a drop-in replacement for E2B with better performance and true VM-level isolation. --- What CubeSandbox Does Spins up KVM-backed microVMs in <60ms using snapshot cloning + CoW memory Provides thousands of concurrent isolated sandboxes per node (<5MB RAM overhead each) Offers E2B SDK compatibility — just change one env var to migrate Enforces kernel-level network isolation via eBPF (CubeVS) Supports single-node and multi-node cluster deployments Enables code execution, shell commands, file ops, browser automation, and RL training --- Requirements x8664 Linux with KVM enabled (bare metal, WSL2, or cloud bare-metal) Not supported on shared VMs that don't allow nested virtualization Check KVM availability: --- Installation Option A: Development VM (WSL2 / no bare metal) Option B: Bare-Metal / Cloud Server Inside the target Linux host (or the dev VM from Option A): This installs cubemastercli and starts the CubeAPI service on port 3000. --- Key CLI: cubemastercli Create a Template from a Docker Image Watch Build Progress List Templates Delete a Template List Running Sandboxes Kill a Sandbox --- Environment Variables --- Python SDK Usage (E2B-Compatible) Install the E2B SDK: Basic Code Execution Run Python with Return Values Shell Command Execution File Operations Install Packages at Runtime Persistent Sandbox (Manual Lifecycle) Concurrent Sandboxes --- Custom Template Creation From a Custom Dockerfile Build and push your image, then create a template: Template with Multiple Exposed Ports --- REST API (CubeAPI) CubeAPI runs on port 3000 and is E2B-compatible. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aradotso/cubesandbox-ai-sandbox