CCA-F Practice Questions — Page 11

Question 101

Scenario: Prompt Engineering and Structured Output -

A CI review agent has a high false-positive rate in one category ("possible performance issue"), which is eroding developer trust in its other, accurate categories.

What is a reasonable immediate action?

A.Temporarily disable the high-false-positive category until its criteria are tightened, preserving trust in the accurate categories.
B.Keep the category but tell developers to ignore its findings.
C.Disable all categories until the one category is fixed.
D.Increase the volume of findings to show the tool is thorough.

Question 102

Open question ↗

Scenario: Prompt Engineering and Structured Output -

You need consistent extraction of informal measurements like "two handfuls of rice" and "a pinch of salt" into a normalized structure. A purely descriptive instruction produces inconsistent results.

Which technique works best here?

A.Few-shot examples that map several informal phrases to the exact desired output structure.
B.A longer prose description of how to handle informal units.
C.Forcing `tool_choice: "any"` to guarantee JSON.
D.Raising temperature so the model is more creative with conversions.

Question 103

Open question ↗

Scenario: Prompt Engineering and Structured Output -

Your extraction service intermittently returns malformed JSONâ€”missing braces and trailing commasâ€”breaking the downstream parser.

What is the most reliable way to eliminate these syntax errors?

A.Use `tool_use` with a JSON Schema so output is guaranteed to be syntactically valid and schema-conformant.
B.Ask the model in the prompt to "output valid JSON only."
C.Post-process the text with a regex that fixes braces and commas.
D.Lower temperature to 0 to avoid formatting mistakes.

Question 104

Open question ↗

Scenario: Prompt Engineering and Structured Output -

After switching to `tool_use` with a strict JSON Schema, output is always valid JSON, yet an invoice extraction still reports `total: 150` when the line items sum to 145.

Why does this happen and what addresses it?

A.This is a semantic error; JSON Schema guarantees structure, not correctnessâ€”add validation checks and retry-with-feedback or self-correction.
B.The schema is malformed; tightening the schema types will fix the arithmetic.
C.The model needs `tool_choice: "any"` to compute the total correctly.
D.Raising `max_tokens` will let the model finish the calculation.

Question 105

Open question ↗

Scenario: Prompt Engineering and Structured Output -

An extraction schema marks `middle_name` as required. Many source documents have no middle name, and the model fabricates plausible-looking values to satisfy the field.

How should the schema be designed?

A.Make the field nullable (`"type": ["string", "null"]`) so the model can return null when the information is absent.
B.Keep it required but add "do not guess" to the prompt.
C.Remove the field from the schema entirely so nothing is reported.
D.Set a default value of "N/A" as a required string.

Question 106

Open question ↗

Scenario: Prompt Engineering and Structured Output -

A ticket classifier uses an enum `["bug", "feature", "docs"]`. Real tickets sometimes don't fit any category, and the model is forced to pick a wrong one, losing information.

How should the enum be extended?

A.Add `"other"` (with a detail string) and `"unclear"` so out-of-scope or ambiguous items are captured honestly.
B.Add 20 more specific categories to cover every possibility.
C.Remove the enum constraint and accept any free-text label.
D.Force the model to always pick the closest of the three categories.

Question 107

Open question ↗

Scenario: Prompt Engineering and Structured Output -

An extraction fails validation because a date is in the wrong format. You plan to retry with feedback. Separately, another extraction fails because the requested value simply isn't present anywhere in the provided document.

For which case will retry-with-feedback help?

A.Retry helps the date-format case (the model can correct format); it won't help when the information is absent from the source.
B.Retry helps both cases equally.
C.Retry helps neither; you must switch models.
D.Retry helps the missing-information case but not the format case.

Question 108

Open question ↗

Scenario: Prompt Engineering and Structured Output -

A team must classify 50,000 archived documents overnight; results are needed by morning and cost matters. No interactive turn-by-turn tool calling is required per document.

Which API choice is most appropriate?

A.The Message Batches API, for ~50% savings within the up-to-24-hour window.
B.The synchronous API, to get each result immediately.
C.The synchronous API with multi-turn tool calling for each document.
D.Split into 50,000 separate real-time requests to avoid the batch window.

Question 109

Open question ↗

Scenario: Prompt Engineering and Structured Output -

A batch of 100 documents is submitted; 95 succeed and 5 fail because they exceeded the context limit. You want to reprocess efficiently without redoing successful work.

How does `custom_id` help?

A.It links each response to its source document, so you can identify and resubmit only the 5 failed documents.
B.It guarantees ordering so failed items are always at the end.
C.It automatically retries failed documents within the same batch.
D.It enables multi-turn tool calling for the failed documents.

Question 110

Open question ↗

Scenario: Prompt Engineering and Structured Output -

You want a self-correction signal in invoice extraction so a downstream system can catch when a stated total doesn't match the line items.

Which schema design enables this?

A.Extract both `stated_total` and `calculated_total` plus a `conflict_detected` flag so discrepancies are surfaced.
B.Extract only `total` and trust the model to compute it correctly.
C.Mark `total` as required so the model must provide it.
D.Use `tool_choice: "any"` to force a numeric total.