Prompt Engineering & Structured Output Quiz - Certifications Claude Architect Foundations

Free practice quiz: Prompt Engineering & Structured Output (Certifications Claude Architect Foundations). 62 multiple-choice questions with answers and explanations. Play online, no sign-up needed.

This Certifications Claude Architect Foundations practice quiz has 62 multiple-choice questions on Prompt Engineering & Structured Output, and 62 of them come with an explanation. Try the 5 sample questions below (tap an option to check yourself), then play the full quiz online. No sign-up needed.

Sample questions

1. What is the most likely reason the instruction 'only report high-confidence findings' did not improve precision?

   1. The confidence guidance lacks report categories, so it uses its own criteria.
   2. The reviewer receives too much code at once and cannot focus on the correct issues.
   3. The model ignores the instruction because it cannot apply confidence thresholds.
   4. The prompt does not require reasoning before the model reports each finding.
   Show answer

   Answer: A. The confidence guidance lacks report categories, so it uses its own criteria.

   Precision improves when the prompt specifies categorical report/skip criteria, not when it asks the model to be conservative or highly confident. Without named categories, the model keeps applying its own judgment to the same noisy findings.

2. What should the examples contain to most effectively fix this routing ambiguity?

   1. Provide a stricter instruction block repeating that the coordinator must read each request carefully before choosing a subagent.
   2. Provide longer subagent descriptions that add more examples of their internal prompts and the capabilities that they offer.
   3. Provide 2-4 examples showing each ambiguous request, the chosen subagent, and a brief reason for choosing it over alternatives.
   4. Provide a numbered list of every possible user phrasing and the exact subagent that should be chosen for each phrasing.
   Show answer

   Answer: C. Provide 2-4 examples showing each ambiguous request, the chosen subagent, and a brief reason for choosing it over alternatives.

   The task calls for creating 2-4 targeted few-shot examples for ambiguous scenarios that show reasoning for why one action was chosen over plausible alternatives. A phrase list only handles the listed phrases, while repeating instructions or expanding subagent descriptions does not demonstrate the routing decision the coordinator should make.

3. What configuration best guarantees that extract_metadata runs first on every document?

   1. Force tool_choice to select the extract_metadata tool and expose only that tool for the initial step.
   2. Set tool_choice to 'any' and add a stronger system prompt telling it to call extract_metadata first.
   3. Set tool_choice to 'auto' and block all other tools until the model returns a text summary.
   4. Remove all extraction tools and use an output_config JSON schema to get metadata as plain text.
   Show answer

   Answer: A. Force tool_choice to select the extract_metadata tool and expose only that tool for the initial step.

   Forcing a specific named tool guarantees the model must call extract_metadata rather than returning text or choosing another tool. With tool_choice 'any', it must call a tool but could choose a different one, while 'auto' may return text instead.

4. Which conclusion about retrying these two failures is most accurate?

   1. Neither failure can be fixed by retries because both are schema syntax errors.
   2. Compliance_owner_email can be fixed by stronger wording in the retry prompt; release_date is impossible without retries.
   3. Release_date is retryable with reformatting advice; compliance_owner_email needs the intranet page or a human.
   4. Both failures can be fixed by retry-with-error feedback because the data is in the provided source files.
   Show answer

   Answer: C. Release_date is retryable with reformatting advice; compliance_owner_email needs the intranet page or a human.

   The release_date failure is a format mismatch in an available source file, so a retry with specific reformatting feedback is likely to succeed. The compliance_owner_email value is absent from the provided repository and exists only in an external intranet page, so retries will not retrieve it. This demonstrates the difference between fixable format errors and missing external information.

5. What is the most effective change for future claim batches?

   1. Create a composite custom_id from claim_number and policy_year, such as claim_number:policy_year, so each request id is unique and results map to source records.
   2. Split the batch into smaller batches of 1,000 claims and continue using claim_number as custom_id to prevent duplicate values from appearing within each smaller batch.
   3. Remove custom_id entirely and match every returned result to the original input array by using the request index from the batch submission file, without relying on ids.
   4. Keep claim_number as custom_id and rely on the order of the returned results because the Batches API preserves input order for requests with the same custom_id.
   Show answer

   Answer: A. Create a composite custom_id from claim_number and policy_year, such as claim_number:policy_year, so each request id is unique and results map to source records.

   The custom_id field is the correlation mechanism for batch request/response pairs, and it must be unique per request to make results unambiguous. A composite key preserves business meaning while guaranteeing uniqueness, whereas result order is not guaranteed and splitting batches does not remove duplicate claim_number values across policy years.

More questions in this quiz

Answers and explanations are in the full quiz.

1. What best explains why developers now ignore the security findings?
2. Which prompt change is most effective for reducing the noise while preserving important findings?
3. What should the team do first to restore developer trust in the automated review?
4. What is the best change to reduce false positives while still detecting real comment-code contradictions?
5. Which change is most likely to improve precision?
6. What is the most likely root cause of the unread security finding?
7. Which prompt change is best?
8. What is the best approach?
9. What should the team add to best teach the agent to choose the right initial tool for these ambiguous cases?
10. The current prompt lists known bad patterns such as SQL string concatenation and hardcoded API keys, but the reviewer misses previously unseen variants that pass tainted data through a new internal helper or read credentials from an unexpected config key. The team wants the reviewer to catch previously unseen variations of these defect classes, not just the entries already in the fixed list. What is the best way to improve that judgment?
11. The schema requires quantity, unit, and shipping_date. Detailed schema instructions still produce null quantity in 14% of documents when sources say 'two dozen' or 'half a pallet', and sometimes the model invents a shipping_date when the document only says 'ships next week'. Downstream systems reject these records. Which change most directly reduces both empty extraction and hallucinated values for these informal and varied documents?
12. What is the most effective way to make the field values and format consistent across runs?
13. The team wants reports to have consistent sections and actionable recommendations. What is the MOST effective first step?
14. The policy includes examples such as string concatenation in SQL queries in Python. The team finds it misses the same unsafe idea when it appears in Go and JavaScript, where the syntax differs and the policy has no matching entry. What change would most help the reviewer transfer the existing safety judgment to these unlisted cases?
15. What is the most effective fix for reducing these fabricated values?
16. The team wants to add few-shot examples to improve the assistant reasoning. Which set of examples would be MOST effective?
17. The configured output should contain location, issue, severity, and suggested fix. The system prompt already lists these fields, but actual review comments often arrive as prose paragraphs with severity missing or buried in a general comment. What is the MOST effective way to achieve consistent finding format?
18. What change most directly reduces these null extractions without causing fabricated values?
19. Which few-shot set would most improve the subagent's handling of these document structure variations?
20. What is the best way to change the few-shot examples to improve generalization while keeping agent behavior grounded?
21. Which change to the few-shot examples is the most effective first step for reducing false positives while preserving the ability to catch new unsafe variants?
22. Which change is the BEST next step to eliminate JSON syntax errors without adding a retry layer?
23. What is the most likely root cause of the downstream rejections?
24. What schema change is most likely to reduce fabrication?
25. Which configuration should the team use to ensure the model always calls one extraction tool?
26. Which tool_choice setting best guarantees structured output when the document type is unknown in advance?
27. What is the most accurate conclusion about these failures?
28. Which schema change best prevents fabricated values while still extracting the field when it is present?
29. Which configuration should the team use?
30. What is the most effective change to reduce downstream rejections without sacrificing schema-valid JSON?
31. What should the retry prompt contain to maximize the chance of a correct extraction?
32. What is the most effective next step?
33. What change would best enable systematic analysis of dismissal patterns?
34. Which classification is most useful for deciding where to invest?
35. What is the best diagnosis and change?
36. What is the most effective field to add to each finding to enable systematic false-positive analysis?
37. What is the most accurate assessment of the developer's proposal?
38. Which conclusion about the retry loop is most accurate?
39. What is the best change to the validation flow?
40. What is the BEST approach for this blocking pre-merge workflow?
41. Which API strategy best matches the latency requirements?
42. Which decision is the best approach?
43. What is the most likely root cause of the 9,000 incomplete requests?
44. Which API strategy best meets these two workload requirements?
45. Which batch submission schedule is the least frequent option that still guarantees the 30-hour data freshness SLA?
46. What is the best design change for future weekly batches?
47. What is the best way to set up the review?
48. Which review architecture is most likely to catch this bug?
49. What is the most effective first step to improve the signal?
50. What is the most likely reason the self-review missed the logged PII?
51. Which change is most effective?
52. What is the most effective review approach?
53. Which workflow best applies multi-pass review?
54. The lead proposes making the self-review language even more emphatic. What is the best assessment of that plan?
55. Which change is most likely to eliminate this kind of contradiction?
56. What is the best way to adjust the integration pass?
57. What is the best change to make routing decisions better calibrated?

More quizzes

• Tool Design & MCP Integration▶ Play
• Agentic Architecture & Orchestration▶ Play
• Claude Code Configuration & Workflows▶ Play

Browse all Claude Architect Foundations quizzes

More to play