Agent loop fundamentals: planning, tools, and stop conditions

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-creators-agent-loop-fundamentals

1. What is the primary risk of running an agent loop without explicit stop conditions?

   1. The agent may continue running indefinitely, consuming compute resources without achieving its goal
   2. The agent will automatically switch to a more efficient algorithm
   3. The agent will require fewer tools to complete tasks
   4. The agent's outputs will become more accurate over time

2. In the plan-act-observe framework, what is the primary purpose of the 'observe' step?

   1. To generate the initial plan
   2. To select which tool to call next
   3. To evaluate whether the current state satisfies the termination criteria
   4. To log all outputs for later analysis by humans

3. Which of the following is NOT typically part of an agent's tool budget definition?

   1. Maximum iterations
   2. The specific output format for results
   3. Maximum wall-clock time
   4. Maximum number of tool calls allowed

4. What does a 'default-deny' policy mean when applied to tool budgets?

   1. Any tool use beyond the defined budget is prohibited unless explicitly approved
   2. Any tool call that exceeds the budget is automatically allowed after a short delay
   3. All tools are permitted by default, and only risky tools require approval
   4. The agent defaults to using the simplest available tool

5. Why is runtime observability critical for agent loops?

   1. It enables developers to understand what the agent is doing and intervene if needed
   2. It allows the agent to learn from its past decisions
   3. It reduces the computational cost of running the agent
   4. It replaces the need for any stop conditions

6. Which of the following represents a valid stop condition for an agent loop?

   1. The agent has achieved the task goal or exhausted its resource budget
   2. The user has asked at least three questions
   3. The agent has called more than 50 tools
   4. The agent has been running for more than 10 minutes

7. What is the recommended initial approach for setting tool budgets?

   1. Default to tight budgets and raise them only when evidence shows more resources are needed
   2. Set budgets based on the most complex possible task
   3. Use the same budget for all agents regardless of task
   4. Start with generous budgets to avoid interruption, then tighten if needed

8. In the plan-act-observe loop, what is the 'plan' component primarily responsible for?

   1. Deciding which tools to call based on current state
   2. Monitoring resource consumption
   3. Generating the final output
   4. Breaking down the overall task into executable steps

9. What specific capability does observability provide that logging alone cannot?

   1. Runtime visibility into the agent's internal state and decision-making process
   2. Automatic correction of agent errors
   3. Guaranteed termination of runaway loops
   4. Permanent storage of all agent outputs

10. A developer designs an agent loop with the following: max 10 iterations, max 100 tool calls, max 5 minutes wall-clock time. What happens when the agent reaches the 5-minute mark but hasn't completed its task?

    1. The agent automatically extends its time budget
    2. The agent pauses and waits for user permission to continue
    3. The agent continues running until it completes the task
    4. The agent stops because the wall-clock stop condition has been met

11. What distinguishes a 'chaotic' agent loop from a 'debuggable' one?

    1. Chaotic loops lack observability and clear stop conditions
    2. Chaotic loops always fail while debuggable ones always succeed
    3. Chaotic loops run faster than debuggable ones
    4. Chaotic loops use more tools than debuggable ones

12. When should a developer consider increasing an agent's initially tight budget?

    1. When the agent completes tasks faster than expected
    2. Immediately upon deployment
    3. When evidence shows the current budget is insufficient for legitimate tasks
    4. At random intervals

13. What is the purpose of assigning an 'owner' for budget exceptions?

    1. To have a person who can authorize exceeding budget limits when justified
    2. To blame someone when the agent overspends
    3. To automatically deny all exception requests
    4. To track which developer wrote the agent code

14. Which statement about runtime observability is correct?

    1. Observability enables intervention during agent execution
    2. Observability is only needed for failed agent runs
    3. Observability and logging are essentially the same thing
    4. Observability automatically fixes agent errors

15. An agent without explicit tool budgets is described as being able to 'spend a month of compute in an afternoon.' What does this illustrate?

    1. The potential for runaway resource consumption without constraints
    2. The agent can complete very complex tasks quickly
    3. The agent is highly efficient
    4. The agent requires no human oversight