Part 2 – Running Your LLM Agent Safely: Hands-on with Docker Sandboxes

Once your LLM agent starts acting on its own — calling tools, writing files, hitting APIs — "where do I run this thing?" becomes a security question, not a deployment footnote. This free, 2-hour hands-on workshop tackles that exact problem with Docker's Dan Ndombe, using Docker Sandboxes (SBX) to give agents an isolated runtime with clear boundaries on what they can touch. What You'll Learn 🔹 How to spin up a sandboxed agent runtime using Docker Sandboxes 🔹 How to assign an agent a real task and watch it execute inside the sandbox 🔹 How to tighten policies and permissions around agent behavior 🔹 A reusable containment pattern for your own agent prototypes Why This Matters An agent with unrestricted system access is an agent with an unrestricted attack surface. A single bad tool call or prompt injection can turn into a real incident. Instead of trusting an agent by default, you contain it by design — and that's exactly what this session walks you through, live and code-along, not slide-by-slide. Who Should Come? Software engineers building agents, DevOps/platform engineers, security-minded developers, and technical leads scaling agentic AI. No prior Docker experience required. Bring your questions — we'll leave time for live Q&A.