Free Lesson
Coding agents: architecture and security intro
150 min
Nov 2, 2026 12:00 PM
By continuing, you agree to Maven's Terms and Privacy Policy.
What you'll learn
Inside the agent loop
Trace one task through the agent loop: tool choice, context retrieval, and a codebase too large for the context window.
Where the risks come from
Secrets in environment or config files can enter context; hidden text in a file or web page can steer the agent.
Safer use and its limits
Sandboxes reduce risk, not guarantee isolation. Limit access, use a throwaway VM/container, and inspect commands.
Why this topic matters
Agents can read repos, run commands, and change files, which makes them useful and risky. We’ll trace one task through an agent loop, then see how secrets and hidden instructions can cause leaks or damage. Sandboxes and approvals reduce risk but cannot guarantee isolation. You’ll practice limiting access, using a throwaway VM or container, and reviewing commands. Participants will get a Moonshot AI API key to try the examples.
You'll learn from
Nikalina Ogorodova
AI/ML Engineer at Hyperskill, making advanced AI concepts accessible.
