What Tools Agents Can Use

End-of-lesson check

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

1. Which sentence best captures the main idea of 'What Tools Agents Can Use'?

   1. Modern agents can use tools — like a browser, an email client, a calculator, a calendar.
   2. Tools and goals are unnecessary for agent design
   3. Agents should always run without limits or oversight
   4. Agents and chatbots are the same thing in every way

2. Which of the following is part of '2026 agent tool list'?

   1. Use the most expensive model regardless of fit
   2. Run unbounded retries on any error
   3. Web browsers (Operator, Claude Computer Use), code editors (Cursor), payment APIs (Stripe), email clients, calendars, document editors, spreadsheets.
   4. Skip every form of evaluation

3. Which of the following is part of 'Three powerful tools agents have'?

   1. Use the most expensive model regardless of fit
   2. Run unbounded retries on any error
   3. Never log what the agent did
   4. Browser (search and click anything)

4. Which of the following is part of 'Review date'?

   1. Always run with no oversight
   2. Avoid taking any actions in the world
   3. Reviewed in 2026. Treat fast-changing product names, prices, availability, and policy details as examples to verify before use.
   4. Run unbounded retries on any error

5. What is 'tool use' in this context?

   1. A way to disable the agent's tools
   2. A core concept covered in What Tools Agents Can Use
   3. A reason to skip all logging
   4. A trick to bypass approvals

6. What is 'API' in this context?

   1. A reason to skip all logging
   2. A way to disable the agent's tools
   3. A core concept covered in What Tools Agents Can Use
   4. A trick to bypass approvals

7. What is 'MCP' in this context?

   1. A way to disable the agent's tools
   2. A reason to skip all logging
   3. A trick to bypass approvals
   4. A core concept covered in What Tools Agents Can Use

8. Which is the best mental model for an agent's 'tools'?

   1. A replacement for the model
   2. Decoration on the prompt
   3. Functions the agent can call, each with a name, description, and input schema
   4. Random extras the model ignores

9. Which is the clearest sign an 'agent' is really just a chatbot in disguise?

   1. It can call a search tool
   2. It can remember last week's conversation
   3. It only produces text and never takes actions
   4. It uses a system prompt

10. What does an 'eval' for an agent measure?

    1. The exact wording of every prompt
    2. The temperature setting
    3. Whether the agent reliably completes a defined task end to end
    4. How polite the model sounds

11. What is the most reliable way to keep an autonomous agent from going off the rails on a long task?

    1. Set a clear goal, a step budget, and review checkpoints
    2. Run it for as many steps as possible without checking in
    3. Trust the model to know when to stop
    4. Disable its tools so it can only think

12. Why does an AI agent need 'tools' such as a browser, calendar, or code runner?

    1. Tools make the model speak more naturally
    2. Tools replace the need for any prompts
    3. Tools let the agent take actions in the world instead of only producing text
    4. Tools shrink the context window

13. Which budget control most directly prevents runaway costs from an agent loop?

    1. A hard cap on steps, tokens, or dollars per task
    2. A longer context window
    3. A bigger model
    4. A friendly system prompt

14. Before letting an agent take a destructive action, what is the safest default?

    1. Require explicit human approval for the specific action
    2. Hide the action from any log
    3. Skip approvals if the user trusts the agent
    4. Approve once and let the agent repeat forever

15. An agent quietly retries a failed payment 50 times overnight. What design principle was missing?

    1. A larger model
    2. More creative prompting
    3. A bigger context window
    4. Bounded retries with human notification on repeated failure