CCAO-F Prompting and Context Engineering Practice Question
You are building a customer support bot using Claude 3.5 Sonnet. You notice the model sometimes hallucinates policy details when the user asks a question not covered in your provided documentation. How should you structure your prompt to minimize this?
⚠ Common exam trap
Candidates often rely on 'be accurate' or 'don't lie' instructions. These are too subjective; the model needs a binary condition to determine when it should stop answering.
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 constraint to the system prompt: 'If the answer is not found in the provided context, state that you do not know and do not attempt to guess.'
Anchoring the model with strict constraints is the most effective way to reduce hallucinations. By explicitly instructing the model to state 'I don't know' rather than guessing, you force a boundary between known context and external knowledge. This approach is essential for enterprise applications where accuracy and safety are paramount, ensuring that the model remains within the provided knowledge base and does not invent plausible-sounding but incorrect information.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Increase the temperature setting to 1.0 to ensure the model explores all possible answers before responding.
Why it's wrong here
Higher temperatures actually increase the likelihood of creative or hallucinatory output. In a customer support context, you want to minimize temperature to ensure deterministic and factual responses, especially when the model lacks the necessary information to provide a verified answer based on the provided corporate documentation.
- ✗
Embed all company policy documents into a single massive system prompt to ensure total coverage.
Why it's wrong here
Overloading the system prompt can lead to performance degradation due to the needle-in-a-haystack problem. Large prompts dilute the model's focus, making it harder to retrieve specific facts accurately. It is better to use Retrieval-Augmented Generation (RAG) to inject only the relevant context for each specific user query.
- ✓
Add a constraint to the system prompt: 'If the answer is not found in the provided context, state that you do not know and do not attempt to guess.'
Why this is correct
This directive creates a clear refusal behavior when the context is insufficient. By explicitly forbidding the model from guessing, you enforce a strict groundedness in the provided information. This prevents the model from relying on its pre-trained internal knowledge, which may be outdated or conflict with specific company policies.
- ✗
Use few-shot prompting to include examples of the model generating creative solutions to unknown problems.
Why it's wrong here
Few-shot prompting with creative examples will encourage the model to hallucinate or invent policies when it lacks data. The goal for a support bot should be to prioritize accuracy and adherence to provided documents over creative problem solving, as corporate liability often rests on providing precise, verified information.
About these practice questions
Courseiva writes every CCAO-F question from scratch — 259 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 CCAO-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 CCAO-F exam.