Production Agents - Q&A Review Bank
32 questions across architecture, security, durability, cost, evaluation and observability. Try each one before revealing the answer.
Learning objectives 45 min
By the end of this page you will be able to:- Recall the key design rules for production agents in each area
- Apply them to short design scenarios
- Explain the trade-offs behind each rule
Prerequisites
- Chapters 1-6 of this module
Architecture and Reliability
Check yourself
0 / 5 answered
- Name the layers of a production agent system.
- Which failures should an agent harness retry, and which not?
- Why do retries require idempotency keys on writes?
- What should happen when a run hits its step or cost bound?
- What makes a human approval gate effective rather than ceremonial?
Security
Check yourself
0 / 6 answered
- State the lethal trifecta and the structural fix.
- Why is prompt-injection resistance in the model not sufficient?
- Name the six injection-resistant design patterns.
- What does CaMeL add to the dual LLM pattern?
- List five risks from the OWASP Top 10 for Agentic Applications.
- How should credentials be handled for an agent that runs code?
Durable Execution
Check yourself
0 / 5 answered
- What does durable execution record, and what does it enable?
- What is the determinism rule?
- Compare Temporal, Restate and DBOS in one line each.
- Is a LangGraph checkpointer durable execution?
- How do you keep a long agent's workflow history manageable?
Cost and Latency
Check yourself
0 / 6 answered
- Why does an agent's input-token cost grow faster than its number of turns?
- Three rules for prompt caching in an agent loop?
- Name four ways to shrink context growth.
- Routing vs cascades?
- Why track cost per successful task rather than cost per request?
- Name four latency levers for agents.
Evaluation and Observability
Check yourself
0 / 10 answered
- What is trajectory evaluation and why is it needed alongside outcome checks?
- How does LLM-as-judge differ for single-turn evaluation and trajectory evaluation?
- Regression vs capability suites?
- How do you validate an LLM judge?
- What do τ-bench's pass^k results show?
- Name four questions to ask about a public agent benchmark score.
- What are the three core span types in the OpenTelemetry GenAI conventions for agents?
- What must you record on every run to tie a regression to a change?
- Walk through debugging 'the agent said it refunded but didn't'.
- Which signals indicate a possible prompt-injection campaign in production?
Last reviewed: 2026-09