CCAR-P Advanced Agentic Architecture Practice Question
A Claude agent performs a multi-step deployment task. Step 3 calls a `deploy_service` tool that returns success, but the subsequent verification step fails because the service is not yet healthy. The agent currently treats any tool success as completion and ends the workflow. Which change best addresses this?
⚠ Common exam trap
The trap here is equating a tool's successful response with the workflow's desired outcome, when asynchronous systems often report acceptance before the effect is observable.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Add a post-tool-use hook that polls the service health endpoint and, if unhealthy, returns a structured error to the model so it can retry or escalate.
The agent conflates tool acceptance with desired end state, so verification must be moved into a deterministic post-tool-use hook that polls health and returns a structured error when the service is unhealthy. That lets the model retry or escalate instead of ending the workflow prematurely. Prompting, return-type changes, and token limits all leave the false-success gap intact.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Instruct the model in the system prompt to always wait 60 seconds and re-check health after every deploy.
Why it's wrong here
Prompting the model to wait and re-check relies on the model reliably following instructions across many turns, which is exactly the probabilistic behavior that failed here. It also hardcodes an arbitrary delay that may be too short or too long. Deterministic verification belongs in code, not in a prompt the model may skip.
- ✗
Increase the agent's max_tokens so it has more room to notice the verification failure in its reasoning.
Why it's wrong here
Token budget does not change the agent's completion criteria; it merely allows longer output. The failure is architectural, in treating tool success as workflow success, not a shortage of reasoning space. More tokens would not make the agent poll health or continue the workflow.
- ✗
Change the `deploy_service` tool to return a boolean instead of a status string so the agent can parse it more easily.
Why it's wrong here
Changing the return type does not address the core problem, which is that the tool reports acceptance rather than eventual health. A boolean success would still be true immediately after the deploy call. Parsing convenience is irrelevant when the signal itself is premature.
- ✓
Add a post-tool-use hook that polls the service health endpoint and, if unhealthy, returns a structured error to the model so it can retry or escalate.
Why this is correct
A post-tool-use hook runs after the deploy call and can independently verify health, converting a false success into a structured signal the model can act on. This closes the gap between the tool's reported success and the actual desired state. It keeps the orchestration deterministic at the verification boundary while letting the model decide the next step.
About these practice questions
Courseiva writes every CCAR-P question from scratch — 262 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Anthropic exam blueprint
This CCAR-P practice question is part of Courseiva's free Anthropic certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the CCAR-P exam.