19 Prompting, Task Execution & Troubleshooting Practice Questions & Answers

1. 1. A marketing manager's prompts keep producing off-target results, even though the prompts seem clear to her. According to Anthropic's prompt-engineering guidance, what is the best way to test whether a prompt is actually clear?

   • A.Count the words; prompts under 50 words are always clearer
   • B.Run the prompt several times; identical outputs prove it is clear
   • C.Append the phrase "be accurate and clear" to the end of the prompt
   • D.Show the prompt to a colleague with minimal context and ask them to follow it; if they are confused, Claude will likely be confused too✓ Answer

   Anthropic's "golden rule of clear prompting" says to show your prompt to a colleague with minimal context on the task: if that person is confused by the instructions, Claude will likely be too, because the model only has the information the prompt supplies. Word counts, output repetition, and generic exhortations like "be clear" do not diagnose missing context or ambiguity.

   Source: Anthropic Prompt Engineering — "Be clear, direct, and detailed": the golden rule of clear prompting; CCAO-F Exam Guide Domain 1Report a problem with this question

2. 2. An operations lead types "write an email about the outage" and gets a bland, unusable draft. Following Anthropic's guidance to treat Claude like a brilliant new employee who lacks context, what is the most effective improvement to the prompt?

   • A.Switch to a more capable model before changing anything in the prompt
   • B.Ask for ten different versions and pick whichever looks best
   • C.Ask Claude to make the email shorter
   • D.Add context: who the audience is, the purpose of the email, the desired tone, and what a successful email should accomplish✓ Answer

   Anthropic advises treating Claude like a talented new employee with no context: it cannot know the audience, purpose, tone, or success criteria unless the prompt states them, so adding that contextual information is what directly fixes an off-target draft. Shortening, generating many variants, or changing models does not supply the missing information that caused the poor result.

   Source: Anthropic Prompt Engineering — "Be clear, direct, and detailed": give Claude contextual information (new-employee framing); CCAO-F Exam Guide Domain 1Report a problem with this question

3. 3. A project manager must produce a quarterly business review that requires analyzing sales data, summarizing competitor activity, and drafting recommendations. Which approach best applies task decomposition?

   • A.Run the same broad prompt several times and merge the answers
   • B.Ask for maximum detail on all sections in a single response
   • C.Write one long prompt containing every requirement so nothing is left out
   • D.Break the work into sequential subtasks — analyze the data, then summarize competitors, then draft recommendations — feeding each result into the next step✓ Answer

   Anthropic's guidance on chaining complex prompts is to break multi-step work into distinct, sequential subtasks so Claude can give each step its full attention, and to pass each step's output into the next prompt; this reduces dropped requirements and makes errors easier to locate and fix. A single mega-prompt or repeated broad prompts divide the model's attention across competing goals.

   Source: Anthropic Prompt Engineering — "Chain complex prompts" (subtask decomposition); CCAO-F Exam Guide Domain 1: apply task decomposition techniquesReport a problem with this question

4. 4. An education coordinator needs Claude to produce weekly program summaries in exactly the same structure every week, but the layout keeps varying. Which prompt change is most likely to produce consistent formatting?

   • A.Shorten the prompt so Claude focuses only on the format
   • B.Include two or three examples of correctly formatted summaries directly in the prompt✓ Answer
   • C.Regenerate the response until one happens to match the desired layout
   • D.Add "be consistent" at the end of the prompt

   Anthropic identifies examples (multishot prompting) as one of the most effective techniques for getting consistent structure: well-chosen examples show Claude the exact target format, which words alone often fail to convey, dramatically improving consistency. Vague adjectives like "consistent," shorter prompts, or hoping a regeneration matches all leave the desired structure unspecified.

   Source: Anthropic Prompt Engineering — "Use examples (multishot prompting)"; CCAO-F Exam Guide Domain 1Report a problem with this question

5. 5. A finance team member asks Claude to recommend one of three vendor contracts that differ in price, service levels, and termination terms. Which prompting technique best fits this complex analysis task?

   • A.Ask only for the final recommendation to keep the output short
   • B.Ask Claude to work through the trade-offs step by step and show its reasoning before giving a recommendation✓ Answer
   • C.Ask the same question with the contracts removed to avoid biasing the answer
   • D.Ask Claude to answer as quickly as possible

   Anthropic's "let Claude think" (chain-of-thought) guidance explains that giving Claude space to reason step by step before answering improves accuracy on complex tasks like multi-factor analysis, and the visible reasoning also lets the human check how the conclusion was reached. Demanding only a fast, final answer removes that reasoning space, and removing the contracts removes the very data the analysis needs.

   Source: Anthropic Prompt Engineering — "Let Claude think (chain of thought prompting)"; CCAO-F Exam Guide Domain 1: adapt strategy for analysis tasksReport a problem with this question

6. 6. A communications specialist pastes a long customer email thread into Claude along with several instructions about how to respond. Claude sometimes treats sentences from the thread as if they were her instructions. What is the best fix?

   • A.Split the thread across several separate messages, one paragraph at a time
   • B.Place the instructions in the middle of the pasted thread so Claude reads them in context
   • C.Summarize the thread from memory instead of pasting it
   • D.Clearly separate and label the pasted thread and the instructions — for example with tags or headers — so Claude can tell content apart from directions✓ Answer

   Anthropic recommends using structural markers such as XML tags to separate different parts of a prompt — instructions, context, and pasted documents — because clear boundaries prevent Claude from mixing up reference material with directions. Burying instructions inside the content worsens the confusion, paraphrasing from memory loses fidelity, and fragmenting the thread does not mark what is instruction versus content.

   Source: Anthropic Prompt Engineering — "Use XML tags to structure your prompts"; CCAO-F Exam Guide Domain 1Report a problem with this question

7. 7. An analyst's draft reads like a casual blog post, but it must brief a chief financial officer. Which prompting technique most directly shifts Claude's tone and perspective for this audience?

   • A.Ask Claude to use longer, more sophisticated sentences
   • B.Assign a role and audience, for example: "You are a communications advisor preparing a briefing for a CFO"✓ Answer
   • C.Remove all headings so the text reads like a formal letter
   • D.Tell Claude to "sound smarter" and regenerate

   Anthropic's role-prompting guidance explains that assigning Claude a specific role and audience shifts its vocabulary, framing, and level of formality to match that perspective — it is described as one of the most powerful ways to tailor tone and domain focus. Longer sentences, vague exhortations, or cosmetic formatting changes do not reorient the response toward what a CFO audience needs.

   Source: Anthropic Prompt Engineering — "Give Claude a role (role prompting)"; CCAO-F Exam Guide Domain 1Report a problem with this question

8. 8. A marketing team wants to use Claude for a brainstorming session on campaign concepts. Which approach best adapts the prompting strategy to a brainstorming task?

   • A.Ask for the single best campaign idea to save time
   • B.Ask for a large number of varied ideas first, deferring evaluation, then narrow down the strongest ones in a follow-up✓ Answer
   • C.Ask Claude to fully evaluate the feasibility of each idea as it generates it
   • D.Give Claude the concept the team already prefers and ask it to confirm it is the best

   Brainstorming is a divergent task, so the effective strategy is to prompt for quantity and variety first and separate idea generation from evaluation — judging or converging too early cuts off the range of options that brainstorming exists to produce. Asking for one idea, evaluating during generation, or seeking confirmation of a pre-chosen concept all collapse the divergent phase and invite agreement bias.

   Source: CCAO-F Exam Guide Domain 1: adapt prompting strategies based on task type (brainstorming); Anthropic Prompt Engineering — clear, direct prompting for the intended taskReport a problem with this question

9. 9. A business analyst wants Claude to assess whether last quarter's sales promotions worked. Which prompt sets up this analysis task most effectively?

   • A.Ask for the analysis to be written in a more formal, executive tone
   • B.Provide the actual sales data and explicit criteria: the metrics to compare, the time periods, and the decision the analysis will inform✓ Answer
   • C.Ask Claude what data it already has about the company's sales
   • D.Ask "analyze our sales performance" and let Claude choose the angle

   For analysis tasks, effective prompting means supplying the data to analyze plus explicit success criteria — which metrics, comparisons, and decisions matter — because Claude has no access to a company's internal figures and a vague request forces it to guess what "performance" means. Asking Claude what data it holds misunderstands that it only sees what the user provides, and tone adjustments do not define the analysis.

   Source: Anthropic Prompt Engineering — "Be clear, direct, and detailed" (specificity and context); CCAO-F Exam Guide Domain 1: adapt strategy for analysis tasksReport a problem with this question

10. 10. A communications professional needs Claude to draft an announcement about a new internal policy. Which drafting approach follows Anthropic's prompting guidance most closely?

    • A.Write the whole draft herself and ask Claude only to approve it
    • B.Keep the prompt open-ended so Claude can be as creative as possible
    • C.Ask for the longest possible draft and plan to cut it down later
    • D.Specify the audience, purpose, tone, approximate length, and format before Claude writes✓ Answer

    For drafting tasks, Anthropic's clear-and-direct guidance is to state the audience, purpose, tone, length, and output format up front, because Claude produces content matched to whatever specifications it is given — unstated requirements become guesses. Open-ended prompts yield generic drafts, using Claude only as an approver wastes its drafting capability, and over-generating to cut later builds in rework.

    Source: Anthropic Prompt Engineering — "Be clear, direct, and detailed": specify audience, tone, length, and format; CCAO-F Exam Guide Domain 1: adapt strategy for drafting tasksReport a problem with this question

11. 11. A project manager uploads several market research reports and asks Claude to summarize key findings for stakeholders. She wants confidence that every claim in the summary actually comes from the reports. What should she add to the prompt?

    • A.Ask for fewer citations so the summary stays readable
    • B.Assume every statement is from the reports because they were uploaded
    • C.Ask Claude to support each claim with a quote or reference to the specific uploaded report it came from✓ Answer
    • D.Ask Claude to write more confidently so the summary sounds authoritative

    Anthropic's guidance for working with documents recommends asking Claude to ground its answers in quotes or references from the provided material, because this both anchors the response in the source text and makes every claim traceable and verifiable — reducing the risk of blended-in general knowledge or fabricated details. Confident tone is not evidence, and uploading documents does not guarantee that every statement derives from them.

    Source: Anthropic Prompt Engineering — long context tips: ground responses in quotes from provided documents; CCAO-F Exam Guide Domain 1: adapt strategy for research tasksReport a problem with this question

12. 12. An HR specialist keeps prompting Claude with "Don't use jargon, don't be boring, don't make it too long" for benefits announcements, with mixed results. Per Anthropic's guidance, how should she rewrite these instructions?

    • A.Ask Claude to explain why its previous drafts were boring
    • B.Repeat each prohibition at both the beginning and the end of the prompt
    • C.Keep the same prohibitions but add "IMPORTANT" before each one
    • D.State what to do instead: for example, "Use plain language a first-year employee would understand, in a friendly tone, in about 150 words"✓ Answer

    Anthropic advises telling Claude what to do instead of only what not to do: affirmative, specific instructions define the target behavior directly, whereas prohibitions leave the desired alternative unstated and force the model to guess. Emphasis markers and repetition amplify the same vague negatives without adding information, and asking for a self-critique does not supply the missing specification.

    Source: Anthropic prompting best practices — provide affirmative instructions (tell Claude what to do rather than what to avoid); CCAO-F Exam Guide Domains 1 & 7Report a problem with this question

13. 13. An operations coordinator's prompt lists ten requirements in one dense paragraph, and Claude consistently misses the requirement buried in the middle. What is the most appropriate diagnosis and fix?

    • A.Results vary naturally, so resend the identical prompt until it works
    • B.The prompt's structure is the problem: restructure it so each requirement is an explicit, numbered item, with critical ones clearly called out✓ Answer
    • C.The model cannot handle that many requirements, so the instruction should be dropped
    • D.Put the entire prompt in capital letters so nothing is overlooked

    Diagnosing an underperforming prompt starts with its structure: Anthropic recommends presenting instructions as explicit, separated items (numbered lists or structured sections) because discrete, clearly delimited requirements are far less likely to be skipped than ones buried mid-paragraph. Dropping the requirement abandons the goal, resending an unchanged prompt repeats the same failure mode, and all-caps text adds emphasis noise without separating the instructions.

    Source: Anthropic Prompt Engineering — "Be clear, direct, and detailed": use numbered lists / structured instructions; CCAO-F Exam Guide Domain 7: diagnose underperforming promptsReport a problem with this question

14. 14. Claude's draft of a product one-pager is factually fine but reads as generic and could describe any company. What is the most effective next step to improve it?

    • A.Reply "make it better" and see what changes
    • B.Start a new chat and paste the same original prompt
    • C.Click retry repeatedly until a more distinctive version appears
    • D.Tell Claude specifically what is missing — the product names, differentiators, metrics, and level of detail expected — and ask for a revision✓ Answer

    Iterating effectively means giving targeted feedback that adds the information the first output lacked: generic output is almost always caused by generic input, so naming the specific products, differentiators, and detail level supplies what Claude needs to produce a distinctive revision. Retrying, vague feedback like "make it better," or restarting with the identical prompt all leave the underlying information gap in place.

    Source: CCAO-F Exam Guide Domain 1 (iterate prompts) & Domain 7 (adjust based on feedback); Anthropic Prompt Engineering — specificity and iterationReport a problem with this question

15. 15. A team's shared prompt for generating weekly reports performs inconsistently, and an associate wants to find out which part of the prompt is causing the problem. What is the soundest troubleshooting approach?

    • A.Adjust one element at a time and compare the outputs, so the effect of each change can be isolated✓ Answer
    • B.Change the wording, the examples, and the format instructions all at once to fix it faster
    • C.Rewrite the entire prompt from scratch on every attempt
    • D.Assume the variation is random and keep regenerating responses

    Prompt refinement is an empirical process: changing a single element per iteration and comparing results lets you attribute any improvement or regression to that specific change, which is how you locate the faulty part of the prompt. Rewriting everything or changing multiple variables at once destroys that attribution, and treating the variation as random gives up on diagnosis entirely.

    Source: Anthropic Prompt Engineering — iterative, empirical prompt development (test one change at a time); CCAO-F Exam Guide Domain 7: identify and diagnose issues with underperforming promptsReport a problem with this question

16. 16. Every member of a marketing team pastes the same brand voice guidelines into Claude at the start of every conversation. Which change best optimizes this workflow on claude.ai?

    • A.Save the guidelines in a separate document and paste them whenever someone remembers
    • B.Shorten the guidelines so pasting them takes less time
    • C.Create a Project with the brand guidelines set as project instructions and knowledge, so every conversation in it starts with that context✓ Answer
    • D.Tell Claude the guidelines once and expect it to apply them in everyone's future chats automatically

    Projects exist precisely for this optimization: project instructions and knowledge are automatically included as context in every conversation within the Project, eliminating repeated manual setup and keeping outputs consistent across the whole team. Telling Claude once does not carry over to other users' separate chats, and the other options merely make the inefficient manual step slightly faster or less reliable.

    Source: support.claude.com — Claude Projects (project instructions and knowledge apply to every chat in the Project); CCAO-F Exam Guide Domain 7: optimize workflowsReport a problem with this question

17. 17. After a very long working session in a single conversation, Claude begins contradicting decisions made earlier and forgetting instructions from the start of the chat. What is the most effective response?

    • A.Ask Claude to "remember everything above" before each new request
    • B.Start a fresh conversation, carrying over a concise summary of the key decisions, instructions, and context✓ Answer
    • C.Wait a while and continue the same conversation later
    • D.Keep going and repeat every earlier instruction inside each new message

    Claude works within a finite context window, and very long conversations degrade because early content competes with or falls outside what the model can attend to; the recommended remedy is to restart with a distilled summary, which restores a clean, focused context containing only what matters. Repeating all instructions bloats the conversation further, "remember everything" cannot expand the context window, and waiting changes nothing because the problem is conversation length, not time.

    Source: support.claude.com / Anthropic docs — context window limits and long-conversation guidance (restart with a summary); CCAO-F Exam Guide Domains 3 & 7Report a problem with this question

18. 18. An HR specialist asks Claude about her company's internal parental-leave policy, and Claude gives a confident, detailed answer that turns out to be invented. What prompt change best prevents this going forward?

    • A.Provide the actual policy document and tell Claude to answer only from it, explicitly saying so if the answer is not in the document✓ Answer
    • B.Instruct Claude to "only say true things" at the start of every question
    • C.Ask Claude to rate its own confidence and accept answers rated as highly confident
    • D.Ask the same question again more firmly until the details stabilize

    Claude has no access to a company's internal documents, so asking about them without providing the source invites hallucination; Anthropic's hallucination-reduction guidance is to ground the model in the actual reference material and give it explicit permission to say the answer is not there, which removes both the knowledge gap and the pressure to fabricate. Generic truthfulness commands, firmer repetition, and self-rated confidence do not supply the missing source and are unreliable accuracy signals.

    Source: Anthropic docs — "Reduce hallucinations": ground responses in provided documents and allow "I don't know"; CCAO-F Exam Guide Domain 7Report a problem with this question

19. 19. A consultant's first prompt to Claude for a client proposal produced a weak draft. She wants to improve the result efficiently. Which iteration practice reflects Anthropic's guidance?

    • A.Send the identical prompt in several parallel chats and combine the best sentences from each
    • B.Treat the first draft as final, since the first response reflects Claude's best effort
    • C.Delete the chat and try an unrelated wording, since the prompt clearly failed
    • D.Review what specifically fell short, then refine the prompt or give follow-up feedback targeting those gaps, repeating until the output meets the need✓ Answer

    Prompting is inherently iterative: the effective loop is to evaluate the output against the goal, identify the specific shortfalls, and feed targeted refinements or feedback back into the next attempt — each cycle adds information that steers Claude closer to the need. Accepting the first draft skips improvement entirely, random rewording discards what the first attempt revealed, and stitching sentences from identical parallel prompts adds effort without adding any new guidance.

    Source: Anthropic Prompt Engineering — iterative refinement of prompts; CCAO-F Exam Guide Domain 1: iterate prompts to improve output qualityReport a problem with this question

Practice questions based on the official Claude Certified Associate – Foundations (CCAO-F) exam guide and Anthropic's public documentation. This is an independent study tool, not affiliated with or endorsed by Anthropic, and does not grant certification. The real exam is 60 questions, 120 minutes, passing at a scaled 720/1000, delivered via Pearson VUE ($99). Official certification page →