CCDV-F questions and answers — Domain 4: Prompt and Context Engineering

6 practice questions on Prompt and Context Engineering, the domain weighted at 11.0% of the CCDV-F exam. Every answer explains why the distractors fail — that is the part that transfers to questions you have not seen before.

Question 1

What does strict: true on a tool definition guarantee?

1. The tool will be called
2. Tool names and inputs conform to their schemas
3. The final response is valid JSON
4. The tool's return value is validated

Answer: B. strict constrains tool calls. Constraining the final assistant response is output_config.format. Neither validates what your tool returns — that is your code's responsibility.

Question 2

Which JSON Schema feature is not supported by structured outputs?

1. enum
2. anyOf
3. pattern
4. additionalProperties

Answer: C. Regex pattern is unsupported, along with allOf, not, if/then/else, dependentSchemas and recursive schemas. Express the format in the field description and validate it yourself after generation.

Question 3

An extraction is schema-valid but the invoice total does not equal the sum of its line items. What does this demonstrate?

1. The schema was written incorrectly
2. Schemas guarantee shape, not truth — business validation is a separate step
3. Structured outputs are unreliable
4. The document needs re-scanning

Answer: B. A schema enforces field presence and types. Cross-field consistency, ranges and referential integrity are business rules your code must check, feeding failures back into a targeted retry.

Question 4

An extraction fails your validation. What should the retry contain?

1. The original prompt, resent unchanged
2. The specific validation error and the offending value
3. A larger set of few-shot examples
4. A request to try harder

Answer: B. A retry only helps if it carries information the first attempt lacked. Name the field, the value and the constraint, cap the attempts, and route persistent failures to human review with the partial result rather than dropping the record.

Question 5

In clear_tool_uses_20250919, what is clear_at_least for?

1. Setting the minimum number of tool uses to keep
2. Ensuring each activation frees enough tokens to justify the cache invalidation it causes
3. Delaying clearing until a token threshold
4. Excluding specific tools from clearing

Answer: B. Clearing invalidates the cached prefix at the point of the edit, so a trivial clear costs a cache write for almost no saving. keep preserves recency, trigger sets the threshold, and exclude_tools protects named tools.

Question 6

Where should a long reference document sit in a prompt, and why?

1. After the question, so the model reads the question first
2. Before the question, tagged and identified — better for grounding and it makes the prefix cacheable
3. Split into fragments interleaved with instructions
4. In the system prompt, always

Answer: B. Long material first with the instruction last is the documented long-context practice, and it has a second benefit: a stable document at the front of the prompt is exactly what prompt caching can reuse. Interleaving destroys both properties.

← Previous: Questions & answers — Domain 3  |  Next: Questions & answers — Domain 5 →