Practice collection
Build architectural judgment one challenge at a time.
Work through five original practice scenarios on trusted boundaries, tool safety, retries, shared state, and side-effect control.
These are original scenario-based practice questions for Agentic Architect Lab. They are not official certification questions.
Challenge 1Authorization boundariesIntermediate
Purchase Order Approval
Examine how an agent should handle an order that needs manager approval before it can be submitted.
A purchase order needs manager approval before an agent can finish the workflow. What should control submission?
Open challengeChallenge 2Tool safetyIntermediate
Untrusted Tool Output
Examine how an agent should use retrieved content that mixes useful facts with hostile instructions.
A retrieved document includes both the answer and an attempt to redirect the agent. What should happen next?
Open challengeChallenge 3Idempotency and paymentsIntermediate
Duplicate Payment Prevention
Examine how an agent should recover when a payment call times out and the previous attempt may already have succeeded.
A payment request timed out, but the processor may already have charged the customer. What should happen before any retry?
Open challengeChallenge 4Multi-agent coordinationFoundational
Shared Agent State
Examine how cooperating agents should coordinate task facts without blurring ownership or access control.
Several agents need the same task facts, but not the same edit rights. How should that coordination work?
Open challengeChallenge 5Reliability and escalationIntermediate
Bounded Retries
Examine how an agent should respond when an operation keeps failing and recovery may or may not be safe.
An operation is failing repeatedly. How should the agent decide between retrying, stopping, and escalating?
Open challenge