Why AI Agents Fail (and How to Catch It Early)

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-builders-agentic-agent-failure-modes-r8a8-teen

1. Which sentence best captures the main idea of 'Why AI Agents Fail (and How to Catch It Early)'?

   1. Agents fail in predictable ways: looping forever, faking success, going off-topic. Knowing the patterns helps you stop them fast.
   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 'Some examples'?

   1. Hide tool calls from the operator
   2. Claude Code edits the same file 8 times trying to fix one bug — that's a loop, kill it.
   3. Never log what the agent did
   4. Disable safety checks for speed

3. Which of the following is part of 'The rule'?

   1. Use the most expensive model regardless of fit
   2. Set a max-step limit and review the agent's last 3 actions before trusting any 'done' message.
   3. Always run with no oversight
   4. Hide tool calls from the operator

4. Which of the following is part of 'You did it!'?

   1. Use the most expensive model regardless of fit
   2. Knowing how things break is half the job.
   3. Hide tool calls from the operator
   4. Approve all actions automatically

5. What is 'agent failure' in this context?

   1. A trick to bypass approvals
   2. A way to disable the agent's tools
   3. A reason to skip all logging
   4. A core concept covered in Why AI Agents Fail (and How to Catch It Early)

6. What is 'infinite loop' in this context?

   1. A trick to bypass approvals
   2. A reason to skip all logging
   3. A way to disable the agent's tools
   4. A core concept covered in Why AI Agents Fail (and How to Catch It Early)

7. What is 'confabulation' in this context?

   1. A way to disable the agent's tools
   2. A reason to skip all logging
   3. A core concept covered in Why AI Agents Fail (and How to Catch It Early)
   4. A trick to bypass approvals

8. Which is a classic agent failure mode rather than a normal model error?

   1. A typo in one sentence
   2. Repeating the same tool call in a loop with no progress
   3. Slow first token
   4. Choosing a slightly off synonym

9. Why does a multi-agent system sometimes outperform a single agent on complex jobs?

   1. Single agents cannot use tools
   2. Specialized roles can divide work and check each other
   3. Multiple agents always cost less
   4. More agents always means more accuracy

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

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

11. 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 shrink the context window
    4. Tools let the agent take actions in the world instead of only producing text

12. Why is logging every tool call an agent makes a baseline requirement?

    1. Logs replace the need for testing
    2. Logs make the model run faster
    3. Logs are needed to debug, audit, and explain agent behavior to users
    4. Logs are only for legal teams

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

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

14. Why is keeping a human in the loop valuable for high-stakes agent actions?

    1. It removes the need for any logging
    2. It catches mistakes before they cause real-world harm
    3. It speeds the agent up
    4. It replaces the model entirely

15. Which of these is the strongest indicator that an agent workflow is ready to scale?

    1. It runs without any logging
    2. It passes a repeatable eval, has cost in budget, and a rollback plan
    3. It used the latest model
    4. It worked once for one user