CCAO-F - Prompting and Task Execution (14% of the exam) - Section 1.3

Iterate prompts to improve output quality.

Improving a result through targeted follow-up: saying specifically what to change, adding an example of what good looks like, or tightening constraints. Candidates should prefer a specific correction to starting again or accepting a weak draft.

specific feedbackexamples of the desired outputtightening constraints

Practice question for this objective

Free samplePrompting and Task Executioneasy

A customer service team lead at an online retailer uses Claude to draft replies to the dozens of delivery-delay complaints her team receives each day. Her prompt already asks for a 'warm, brief and professional' tone, but the drafts vary widely in tone and structure, so agents rewrite most of them. The team has a set of approved replies that customers have responded well to. What is the best change to her approach?

AAdd more tone words, such as friendly, reassuring and polished, to the prompt
BSwitch every draft to the most capable model so the tone is read correctly
CRegenerate each draft with the same prompt until one matches the house style
DAdd two approved replies, with customer details removed, as examples to follow Correct

When descriptions of style produce inconsistent results, include a short example of good output for Claude to follow. Words like warm and professional mean different things to different readers, so Claude's interpretation drifts from draft to draft. A concrete example of approved output fixes the target in tone, length and structure at once, which gives consistent drafts across a high volume of similar requests.

Why A is wrong: More adjectives feel like a more precise instruction. But tone words are open to interpretation, and piling on more of them does not show Claude what this team's replies actually look like.

Why B is wrong: A stronger model can seem the answer to inconsistent output. But for dozens of short routine replies it adds cost and delay, and the gap is the missing example, not reasoning power.

Why C is wrong: Rerunning sometimes lands on a good draft, so it can look workable. But at dozens of replies a day it wastes time, and an unchanged prompt gives Claude nothing new to aim at.

Why D is correct: The team already knows what a good reply looks like. Showing Claude two approved examples conveys the tone and structure far more precisely than adjectives, and removing customer details keeps personal data out of the prompt.

See more CCAO-F practice questions, answers explained.

Exam traps in Prompting and Task Execution

Answers that look right on this material and are not. Each one is a distractor from a different question in the CCAO-F bank for this domain.

Ask Claude to continue from its earlier outcomes and draft the six lesson plans that follow on from them.
Why it is wrong: Continuing in the same chat feels natural, but Claude's earlier outcomes are not the approved ones. The lessons would be built on two outcomes the school has already replaced.
Ask for a shorter article, on the basis that shorter text will naturally read as less formal and less stiff.
Why it is wrong: Shorter text can feel lighter, so this is a reasonable guess. Length does not set the voice, though; a short article can be just as formal, and the prompt would still not tell Claude how it should sound.
Send a separate follow-up for each remaining term, naming them one at a time as she spots them in the draft.
Why it is wrong: This is specific and would fix each term she names. But it depends on her catching every instance and takes several rounds, whereas stating the rule once lets Claude find them all.

Examworthy is not affiliated with or endorsed by Anthropic. Original, blueprint-aligned practice material only.