Agent Safety: Sandboxes and Human-in-the-Loop

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-agentic-safety-sandbox-builders

1. What is the core idea behind "Agent Safety: Sandboxes and Human-in-the-Loop"?

   1. Giving an AI the keys to your computer is a big deal. Learn the two simplest ways to keep an agent safe: wall it off from things it shouldn't touch, and put a human in the decision path.
   2. Recover attribution for untagged historical calls.
   3. Including correlation IDs across distributed calls
   4. Choosing tools that match keywords in user requests

2. Which term best describes a foundational idea in "Agent Safety: Sandboxes and Human-in-the-Loop"?

   1. human-in-the-loop
   2. sandbox
   3. least privilege
   4. approval gate

3. A learner studying Agent Safety: Sandboxes and Human-in-the-Loop would need to understand which concept?

   1. sandbox
   2. least privilege
   3. human-in-the-loop
   4. approval gate

4. Which of these is directly relevant to Agent Safety: Sandboxes and Human-in-the-Loop?

   1. sandbox
   2. human-in-the-loop
   3. approval gate
   4. least privilege

5. Which of the following is a key point about Agent Safety: Sandboxes and Human-in-the-Loop?

   1. Vercel Sandbox — Firecracker microVMs for running untrusted code.
   2. Docker containers with bind-mounted folders only (classic and free).
   3. Anthropic's Claude Code with scoped permissions (per-directory, per-tool).
   4. Browser-only agents (Browser Use, Operator) — can't touch your filesystem.

6. Which of these does NOT belong in a discussion of Agent Safety: Sandboxes and Human-in-the-Loop?

   1. Vercel Sandbox — Firecracker microVMs for running untrusted code.
   2. Anthropic's Claude Code with scoped permissions (per-directory, per-tool).
   3. Recover attribution for untagged historical calls.
   4. Docker containers with bind-mounted folders only (classic and free).

7. What is the key insight about "Assume the agent will mess up" in the context of Agent Safety: Sandboxes and Human-in-the-Loop?

   1. Recover attribution for untagged historical calls.
   2. Including correlation IDs across distributed calls
   3. Not because AI is evil, but because AI is wrong some of the time.
   4. Choosing tools that match keywords in user requests

8. What is the key insight about "The rule of three" in the context of Agent Safety: Sandboxes and Human-in-the-Loop?

   1. Recover attribution for untagged historical calls.
   2. Including correlation IDs across distributed calls
   3. Choosing tools that match keywords in user requests
   4. Three questions before giving an agent any permission: (1) What's the worst that could happen? (2) How would I notice? (…

9. What is the key insight about "Never give agents your master password" in the context of Agent Safety: Sandboxes and Human-in-the-Loop?

   1. Use API keys with scoped permissions. Use app-specific passwords. Revoke when a task finishes.
   2. Recover attribution for untagged historical calls.
   3. Including correlation IDs across distributed calls
   4. Choosing tools that match keywords in user requests

10. Which statement accurately describes an aspect of Agent Safety: Sandboxes and Human-in-the-Loop?

    1. Recover attribution for untagged historical calls.
    2. An agent with filesystem access can delete your thesis. An agent with email access can send 'you're fired' to your entire team.
    3. Including correlation IDs across distributed calls
    4. Choosing tools that match keywords in user requests

11. What does working with Agent Safety: Sandboxes and Human-in-the-Loop typically involve?

    1. Recover attribution for untagged historical calls.
    2. Including correlation IDs across distributed calls
    3. A sandbox is a walled space where the agent can only see and touch specific things. Outside the wall, the agent has no power.
    4. Choosing tools that match keywords in user requests

12. Which of the following is true about Agent Safety: Sandboxes and Human-in-the-Loop?

    1. Recover attribution for untagged historical calls.
    2. Including correlation IDs across distributed calls
    3. Choosing tools that match keywords in user requests
    4. For any action that's destructive (delete, send, pay, push), require a human to approve before the agent proceeds. Yes, it slows things down.

13. Which best describes the scope of "Agent Safety: Sandboxes and Human-in-the-Loop"?

    1. It focuses on Giving an AI the keys to your computer is a big deal. Learn the two simplest ways to keep an agent s
    2. It is unrelated to agentic workflows
    3. It applies only to the opposite beginner tier
    4. It was deprecated in 2024 and no longer relevant

14. Which section heading best belongs in a lesson about Agent Safety: Sandboxes and Human-in-the-Loop?

    1. Recover attribution for untagged historical calls.
    2. Defense one: sandbox
    3. Including correlation IDs across distributed calls
    4. Choosing tools that match keywords in user requests

15. Which section heading best belongs in a lesson about Agent Safety: Sandboxes and Human-in-the-Loop?

    1. Recover attribution for untagged historical calls.
    2. Including correlation IDs across distributed calls
    3. Defense two: human-in-the-loop
    4. Choosing tools that match keywords in user requests