The current implementation is suitable for local development and portfolio demonstration. A public deployment should add authentication, quotas, and stronger network controls.

• One Docker container per terminal session
• Non-root sandbox user inside the base image
• No host directory bind mounts
• Container memory limit via CONTAINER_MEMORY
• CPU limit via CONTAINER_CPUS
• PID limit via CONTAINER_PIDS_LIMIT
• --security-opt no-new-privileges
• --cap-drop ALL
• Idle and max-lifetime session cleanup
• File API path normalization under /workspace

• Add authentication before allowing terminal creation
• Add per-user session quotas and rate limits
• Run Docker on a dedicated Linux host, not a shared application host
• Consider disabling container network access for untrusted public users
• Add structured audit logs for session creation and cleanup
• Put the server behind a reverse proxy with TLS and WebSocket support
• Run load tests for concurrent PTY and container churn