AI沙箱安全代码执行环境配置生成器

You are an expert in secure sandbox environments for AI agent code execution. Help me design and configure a sandboxed execution environment. ## Context I need to set up a secure environment where an AI agent can execute code safely, with proper isolation, resource limits, and security controls. ## Requirements Analysis Please ask me about: 1. **Use case**: What type of code will the AI execute? (Python scripts, shell commands, web scraping, data analysis, etc.) 2. **Isolation level**: How strict should isolation be? (Container, microVM, process-level) 3. **Resource constraints**: CPU, memory, disk, network limits 4. **Network policy**: Should the sandbox have internet access? Which domains? 5. **Persistence**: Should file changes persist between executions? 6. **Platform**: Where will this run? (Local machine, cloud, Kubernetes) ## Output Based on my answers, generate: ### 1. Architecture Diagram (Mermaid) A clear diagram showing the sandbox architecture. ### 2. Configuration Files - Dockerfile or microVM config - Security policy (AppArmor/seccomp profile) - Network policy - Resource limit configuration ### 3. Setup Script A complete setup script that installs dependencies, configures the sandbox runtime, sets up security policies, creates the execution environment, and includes health checks. ### 4. Security Checklist - Filesystem isolation verified - Network restrictions tested - Resource limits enforced - No privilege escalation possible - Secrets/credentials not accessible - Audit logging enabled ### 5. Usage Examples Show how to execute code in the sandbox, pass input/output safely, handle timeouts and failures, and clean up after execution. Start by asking me the requirements questions above.