Which prompting pattern is best suited for a task that has multiple stages where you do not want the model to skip ahead to a single guess?
AFew-shot examples
BStep-by-step reasoning
CExplicit output formatting
DAdding more role context
Explanation: Step-by-step reasoning is designed exactly for multi-stage thinking — it forces the model to walk through each stage rather than collapse to one answer. Few-shot is tempting because it 'teaches' the model, but it targets style and tone, not the depth of reasoning.
Q2 / 7ApplyLO-S2-1, LO-S2-2
A Data and ERP Analyst is writing a prompt to summarize last quarter's purchase orders. The output will be pasted directly into a spreadsheet for downstream use. Which pattern should the analyst apply as the primary fit?
AStep-by-step reasoning, so the analyst can audit the logic
BFew-shot examples of past spreadsheets the analyst built
CExplicit output formatting (e.g., a table with named columns)
DA longer role description for the model
Explanation: When the answer will be shared, scanned, or pasted into another tool, output formatting is the lever that locks structure. Step-by-step would produce useful reasoning but in prose — still a manual cleanup job before the spreadsheet, which defeats the goal.
Q3 / 7AnalyzeLO-S2-2, LO-S2-3
A Mold and Manufacturing Engineer asks the model about a recurring flash defect and gets a generic textbook list of causes. What is the best diagnosis of why the response missed the mark, and the correct first refinement move?
AThe model is weak; switch to a different model and re-ask the same prompt
CThe output format was wrong — tighten the format to a numbered list
DThe answer is fine — ask a follow-up about clamp tonnage to keep going
Explanation: Generic in, generic out — diagnosing the cause as 'missing context' points directly to the Add Context refinement move. Option D is the classic Follow-up-instead-of-Refine pitfall: piling a new question on top of a flawed turn just propagates the flaw.
Q4 / 7ApplyLO-S2-2
A BioTech CEO wants the model to write a one-paragraph board briefing in a very specific tone she has used before. She has two well-written sample briefings on hand. Which prompting pattern should she apply, and why?
AStep-by-step reasoning, because board briefings have multiple stages
BFew-shot examples, because two real samples teach tone better than describing it
COutput formatting only, because the format will determine the tone
DShe should describe the tone in detail and avoid showing examples
Explanation: Tone and cadence are notoriously hard to specify in the abstract — two genuine examples efficiently show the model the target style. Describing the tone in detail (option D) is the common misstep; it produces verbose style instructions that the model interprets loosely compared to concrete samples.
Q5 / 7ApplyLO-S2-1, LO-S2-2
An Agile Program Manager is planning a sprint and also needs the output as a clean checklist she can paste into the team wiki. Which prompt construction best fits the situation?
AA prompt that uses only few-shot examples of past sprints
BA prompt that picks one pattern strictly, since patterns should never be combined
CA prompt that layers step-by-step reasoning for the planning stages with explicit output formatting for the checklist
DA long free-form prompt with no pattern at all
Explanation: Patterns are not mutually exclusive — sprint planning needs multi-stage reasoning, and a wiki-ready checklist needs locked structure, so layering both into one prompt is the right move. Option B reflects the misconception that you must choose exactly one pattern per prompt.
Q6 / 7EvaluateLO-S2-3
You ask the model for a project status email. The response is on-topic, the format matches what you wanted, and it meets your written stop condition of 'under 120 words, three bullets, neutral tone.' What is the correct next step under the Accept / Refine / Follow-up rule?
ARefine once more — every output can be improved 5%
BFollow-up with 'now make it better'
CAccept — ship it and stop the loop
DStart over with a brand-new prompt to verify the result
Explanation: When the response clears your pre-set stop condition, the rule says Accept — that is the entire point of writing a stop condition. Choosing 'Refine once more' is the perfectionism pitfall: without a defined bar, the loop never ends.
Q7 / 7EvaluateLO-S2-3
A BioTech CEO prompts: 'Summarize our Phase 2 trial results.' The model returns a 400-word wall of text mixing endpoints, statistics, and dosing notes. It is factually correct but not board-ready. What is the most appropriate response under the decision rule?
AAccept — it is technically correct, so it is good enough
BFollow-up by asking 'what about safety data?' on top of the same response
CRefine by layering targeted fixes: add audience, tighten format, set tone
DRefine by changing five things at once to maximize improvement
Explanation: The answer is on-topic but misses the mark on audience, format, and tone — that is the textbook Refine case, and layering one or two targeted fixes is the prescribed move. Option D violates the 'change one or two levers, not five' rule, which makes it impossible to learn what actually fixed the response.