CCDV-F Agents and the Agent SDK Practice Question
A developer is implementing a 'Human-in-the-loop' (HITL) pattern for an agent that performs financial transactions. Which TWO strategies are most important for maintaining the security and reliability of this agentic workflow?
⚠ Common exam trap
Candidates often focus only on the approval step, forgetting that providing the human's rejection feedback back to the agent is critical for enabling the model to learn and correct its path.
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
✓
Providing the human with a clear diff of the proposed transaction data.
HITL is essential for high-stakes tasks where model errors could have significant real-world consequences. By introducing a manual approval step, developers can ensure that the agent remains within its defined boundaries. Clear communication of the agent's intent and a robust mechanism for human rejection are key pillars of this safety architecture.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Providing the human with a clear diff of the proposed transaction data.
Why this is correct
A clear visual representation of what the agent intends to do—such as the amount, recipient, and currency—allows the human to quickly verify the action. Without this transparency, the 'approval' becomes a blind click, defeating the purpose of having a human reviewer to prevent hallucinated or malicious tool arguments.
- ✗
Allowing the agent to bypass human approval if its confidence score is high.
Why it's wrong here
Model confidence scores are often poorly calibrated and do not guarantee accuracy. For financial transactions, bypassing human oversight based on a model-generated probability is a major security risk that could lead to unauthorized fund transfers if the model is overconfident in a hallucinated or incorrect reasoning path.
- ✓
Feeding the human's rejection feedback directly back into the agent's context.
Why this is correct
When a human rejects an action, the agent needs to know why so it can correct its course. By including the rejection reason in the conversation history, the agent can refine its reasoning and attempt a different approach that aligns with the user's constraints and the security policies in place.
- ✗
Obfuscating the tool names from the human to prevent cognitive overload.
Why it's wrong here
Obfuscation is counterproductive in a security context. The human reviewer needs full visibility into which tools are being called and what parameters are being passed. Hiding technical details makes it harder for the reviewer to identify subtle errors or deviations from the expected behavior of the financial agent.
- ✗
Using a shorter system prompt to ensure the agent acts more autonomously.
Why it's wrong here
Autonomy is generally the opposite of what is desired in a high-stakes HITL financial workflow. A more detailed system prompt that strictly defines the limits of the agent's power and explicitly mentions the requirement for human approval is actually more effective for maintaining control and ensuring the agent behaves predictably.
About these practice questions
One of 257 original CCDV-F practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. 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 CCDV-F 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 CCDV-F exam.