Week 4 Quiz — Prompting II — Meta-Prompting & Structured Prompts

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Covers: meta-prompting (Skill 4) · the nine structured-prompt components and what each controls (Skill 5) · over-engineering · classic misconceptions about Role, Examples, and Constraints
Format: 10 auto-graded items (multiple-choice, multiple-answer, matching, true/false) · 10 points (1 each) · allowed attempts: 1 · No AI on this quiz.

This is the human-readable quiz with its vetted answer key and one-line feedback. The import-ready Classic QTI 1.2 is in F-quiz-week-04-qti.xml (generated by a validated Python script — parses with 10 items, every single-answer item exactly one correct). Reminder: AI is not permitted on quizzes.

Questions, key, and feedback

Q1 (MC). Meta-prompting is best described as —
- A. using capital letters and bold text to emphasize the most important part of a prompt
- B. asking the AI to help write or refine a prompt by posing clarifying questions one at a time ✅
- C. copying a prompt from an online library and pasting it unchanged
- D. writing the same prompt in multiple AI tools to compare results
Feedback: Meta-prompting means using the AI to help BUILD the prompt — you ask it to interview you ("ask me clarifying questions one at a time; then return a Markdown prompt") and it surfaces what information it needs to do a great job.

Q2 (MC). You want to use meta-prompting to draft a reusable prompt. Which opening instruction best starts the technique?
- A. "Write me the best possible prompt for drafting a study schedule."
- B. "I need a reusable prompt for drafting a study schedule. Ask me clarifying questions one at a time — when you have what you need, return a Markdown prompt I can copy." ✅
- C. "Give me tips for writing a better prompt for a study schedule."
- D. "You are a prompt engineer. Tell me the ideal study-schedule prompt."
Feedback: The correct form asks the AI to interview you one question at a time and then return a Markdown template. Option A asks the AI to skip the interview and guess; options C and D get advice, not a usable reusable prompt.

Q3 (Matching). Match each structured-prompt component to the question it answers.

Feedback: Context = the world the AI steps into; Goal = the primary task; Audience = who reads it; Constraints = what to avoid/prohibit; Evaluation = the built-in self-check. Each controls a different lever — a prompt missing any of these leaves that lever unset.

Q4 (MC). A student writes "You are a licensed attorney" in their prompt. What does the Role component actually do?
- A. It gives the AI access to verified legal databases and makes its legal information trustworthy
- B. It shapes the AI's style and framing, but does not make its output factually accurate — you still need to verify ✅
- C. It causes the AI to refuse to answer anything outside legal topics
- D. It unlocks a special legal-reasoning mode in the model
Feedback: Role changes style and framing, not accuracy. The AI still generates plausible text — it doesn't retrieve verified legal facts. Writing "you are a lawyer" changes how the output sounds, not whether its legal content is correct. Always verify.

Q5 (MC). Which scenario best illustrates "over-engineering" a prompt?
- A. Adding an Audience component to a prompt that will be read by non-experts
- B. Including a Constraints section in a template you plan to reuse 50 times
- C. Writing a 500-word prompt full of contradictory instructions to generate a three-sentence bio ✅
- D. Asking for a specific output format (a numbered list) for a task that will produce a list
Feedback: Over-engineering = adding so many (or conflicting) components that the prompt works worse. A 500-word setup for a 3-sentence output is the classic case. Options A, B, and D are appropriate uses of components that change the output.

Q6 (Multiple-answer). Which of the following are TRUE about the nine structured-prompt components? Select all that apply.
- A. You do not need all nine components for every prompt — use only the ones that change the output ✅
- B. The Evaluation component is optional and almost never improves output
- C. The Examples component shows the AI the style or format you want, while the Constraints component tells it what to avoid ✅
- D. The Role component affects the tone and framing of the output, but does not guarantee factual accuracy ✅
- E. A prompt with more components is always better than a shorter one
Feedback: True: A (not all nine every time), C (Examples = show style; Constraints = what to avoid), D (Role shapes style, not accuracy). B is false — Evaluation is valuable for high-stakes outputs. E is false — more components can create contradictions and degrade results.

Q7 (MC). A student's prompt says: "Write a blog post about renewable energy. Make it educational. Be sure it's persuasive. Don't be preachy. Target experts. Target beginners." What is the main problem?
- A. The prompt is too short to produce a useful output
- B. The prompt is missing a Role component
- C. The prompt contains contradictory instructions that will confuse the output ✅
- D. The prompt does not specify a word count
Feedback: "Educational + persuasive + not preachy + expert audience + beginner audience" are contradictory. The AI will try to satisfy all instructions and produce something muddled. The fix: pick a consistent goal and audience before adding voice/constraint details.

Q8 (True / False). Asking for a Markdown-formatted prompt when you meta-prompt is mainly a style preference — it has no practical effect on the result.
- True
- False ✅
Feedback: False. Requesting Markdown output makes the components visible and structured, makes the template easy to copy and reuse, and often reveals gaps you can then fill. It's a practical step, not just cosmetic.

Q9 (MC). Here is a prompt fragment: "Before returning the draft, verify: is this under 200 words? Does it use plain language a non-expert could follow? Does it avoid bullet points?" Which structured-prompt component does this represent?
- A. Constraints
- B. Voice/Format
- C. Evaluation ✅
- D. Goal
Feedback: The Evaluation component is the built-in self-check — it asks the AI to test its own output against success criteria before returning it. Constraints say what NOT to do in the content; Evaluation says what to CHECK before delivering it. These are different jobs.

Q10 (MC). A student is writing a prompt to get help drafting a short email to a professor asking for an extension. Which set of components is most essential — the others are optional for a task this small?
- A. Role · Data/Logic · Examples · Evaluation
- B. Goal · Audience · Constraints ✅
- C. Context · Voice/Format · Examples · Evaluation
- D. Role · Constraints · Data/Logic · Voice/Format
Feedback: For a short, one-off task like this, the essentials are Goal (ask for an extension), Audience (a professor — formal tone implied), and Constraints (brief, respectful, don't beg). The heavier components (multiple Examples, detailed Evaluation, Data/Logic) are overkill here — they add complexity without improving a short email.

Answer key (quick reference)

Blueprint & item-bank note

All 10 items are tagged course=AI101 · week=4 · objective=2 and deposited into the item bank. Distractors target the week's classic misconceptions: "Role = expertise/accuracy," "more components = better prompt," "Examples and Constraints are interchangeable," "Markdown is cosmetic."

Quality gate (self-checked)

• Structure: 10 items, 1 point each; types = 6 MC + 1 matching + 1 multiple-answer + 1 true/false + 1 MC ("what's-the-problem?" scenario). ≥1 matching (Q3: component→what it controls). ≥1 "what's the prompting fix?" scenario (Q7, Q10). Requirements met.
• Single-answer integrity: every MC and the true/false item has exactly one correct option; the matching item pairs one-to-one; the multiple-answer item keys A, C, D.
• Product-accuracy gate: PASS. No specific product features, plan tiers, or version-dependent claims. The nine components are a general prompting framework, not specific to any one AI vendor. All tool references (ChatGPT, Claude, Gemini, Copilot) are named factually; no fabricated features.
• QTI parse confirmation: F-quiz-week-04-qti.xml parses as imsqti_xmlv1p2 with 10 items; each single-answer item has exactly one scoring condition set to SCORE=100; the matching item uses partial-credit blocks.

Canvas placement block

canvas_object    = Quizzes::Quiz
title            = "Week 4 Quiz — Prompting II — Meta-Prompting & Structured Prompts"
assignment_group = "Quizzes"
points_possible  = 10
grading_type     = points
available_from_offset_days = 0
due_offset_days  = 6
published        = true
allowed_attempts = 1
shuffle_answers  = true
ai_permitted     = false
provenance       = "~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com"