- Published:
- Last updated:
AI Agents and Tool Use
- Authors
- Name
- Lucian Oprea
- @LucianDSA_
00:30:00
Filter by difficulty
34 of 34 questions shown
Choosing the Right Control Model
⏷ 1. What is an AI agent, and how is it different from a chatbot or fixed workflow?
⏷ 2. When should you prefer a deterministic workflow over an agent?
⏷ 3. How does a basic agent loop work?
⏷ 4. How does an agent use a tool without executing it inside the model?
⏷ 5. What makes an agent tool safe and useful?
Free preview complete
You’ve reached the end of the free preview
Get every remaining question and complete answer, plus progress tracking across the full Interview Question Library.
- 29 more questions and complete answers in this topic
- Full access to every interview topic
- Progress tracking and question flags
- New questions and improvements during your subscription
Full access from
$12/month
No long-term commitment. Cancel whenever you want.
⏷ 6. How do you bound an agent run and decide that it is finished?
Tool Design and Orchestration
⏷ 7. How should you design tool names, descriptions, and schemas?
⏷ 8. How do you help an agent select the right tool?
⏷ 9. How should a runtime validate tool arguments, results, and errors?
⏷ 10. When can tool calls run in parallel, and when must they be sequential?
⏷ 11. How do manager agents, agents-as-tools, and handoffs differ?
⏷ 12. When is one agent better than a multi-agent system?
⏷ 13. What context and authority should cross an agent handoff?
State, Memory, and Durable Execution
⏷ 14. How do working state, conversation memory, and knowledge differ?
⏷ 15. How do you control context growth during a long agent run?
⏷ 16. How do checkpoints make an agent run resumable?
⏷ 17. How do you design long-running agent work with status, cancellation, and resume?
⏷ 18. How do idempotency and retries protect agent side effects?
⏷ 19. How do you prevent concurrent workers or agents from corrupting shared state?
Reliability, Safety, and Human Control
⏷ 20. How should timeouts, retries, and backoff work for agent tools?
⏷ 21. Which agent actions should require human approval?
⏷ 22. How do you apply least privilege to agent tools?
⏷ 23. How do prompt injection and malicious tool output affect agentic systems?
⏷ 24. When and how should agent execution be sandboxed?
⏷ 25. How do you handle partial failure and compensation across multiple tools?
⏷ 26. How do you detect and stop loops, repeated actions, and no progress?
⏷ 27. When should an agent ask for user input or escalate to a human?
Observability, Evaluation, and Protocols
⏷ 28. What should you trace in an agent run?
⏷ 29. How do you evaluate an agentic system end to end?
⏷ 30. What is MCP, and what problems does it not solve?
Identity, Durable State, and External Events
⏷ 31. How should an agent use delegated authorization across tools?
⏷ 32. How should an agent decide what to write to long-term memory?
⏷ 33. When should agent orchestration use an explicit state machine?
⏷ 34. How should an agent handle tools that finish asynchronously?