Chapter 11 Quiz: Prompt Engineering Patterns
Question 1
What is the fundamental difference between a prompt and a prompt pattern?
A) Patterns are longer than prompts B) Patterns use advanced AI techniques; prompts do not C) Patterns are reusable templates with variable placeholders that work for any instance of a task type; prompts are specific to one task instance D) Patterns are stored on the AI platform; prompts are stored locally
Show Answer
**C** — A prompt is specific to one task. A pattern is a generalized template that can be applied to any instance of a task type by filling in [bracket variables]. The key characteristics of a pattern are that it's generalizable (works for many instances), parameterized (has clear variables), and documented (stored and findable). The same structure applied to a single task without any generalization is just a prompt.Question 2
You need to sort 200 customer support tickets into five categories for routing. Which of the 15 patterns is the best primary match?
A) Pattern 3: The Analyzer B) Pattern 9: The Extractor C) Pattern 10: The Classifier D) Pattern 14: The Checker
Show Answer
**C** — The Classifier is designed specifically for sorting items from a list into defined categories. The Analyzer evaluates content against criteria (not the same as sorting into buckets). The Extractor pulls structured information from unstructured text (not sorting). The Checker looks for issues in content. The Classifier pattern includes the category definitions, classification criteria, and output format specification that make systematic sorting reliable.Question 3
Which of the following is NOT one of the three properties of a good reusable pattern?
A) Generalizable — works for many instances of the task type B) Parameterized — uses [bracket variables] for elements that change C) Optimized — produces the longest possible output D) Documented — stored with a use case description and example
Show Answer
**C** — The three properties of a good reusable pattern are generalizable, parameterized, and documented. Length optimization is not a pattern property — some patterns should produce short outputs (the Classifier) and some should produce long ones (the Scaffolder). The goal is producing the right output for the task, not the longest output.Question 4
Alex wants to generate product naming options for a new kitchen product. She uses Pattern 4 (The Generator) and gets 8 names that are all variations of the same style — punny, short, lifestyle-adjacent. What is the most likely cause?
A) The Generator pattern is not suitable for naming tasks B) She did not include diversity requirements in her prompt — no constraints on range of name types C) She should have used the Brainstormer instead D) She requested too many options — 8 is too many for consistent quality
Show Answer
**B** — The Generator pattern produces diverse results only when you specify diversity requirements. Without instructions like "cover both functional and evocative names," "include at least 2 unexpected or unconventional options," or "range from professional to playful," the model defaults to the most common style for the task type. Diversity constraints are the Generator's most important parameters — this is emphasized explicitly in the chapter.Question 5
Pattern 5 (The Critic/Reviewer) specifies a [ROLE] variable. Why does the role matter for a review prompt?
A) Different roles have access to different information B) The role establishes an evaluative perspective with implicit quality standards — a security researcher and an executive editor will critique the same code or content very differently C) Roles make the output longer, which provides more feedback D) The role variable is optional and primarily affects the tone of the feedback
Show Answer
**B** — The role in a Critic/Reviewer pattern is not stylistic — it determines which evaluation framework and quality standards are applied. "Review as a potential customer" focuses on buying signals and objections. "Review as a senior legal counsel" focuses on liability risks. "Review as a skeptical journalist" focuses on unsubstantiated claims. The role selection is the most important decision in the Critic pattern because it determines what gets flagged.Question 6
What is the key rule in Pattern 9 (The Extractor) that prevents the model from hallucinating extracted information?
A) Limiting the number of extracted fields to under 10 B) Specifying "only include information explicitly stated in the source" and defining what to write when a field is not present C) Using double brackets [[VARIABLE]] instead of single brackets for required fields D) Running the extraction twice and comparing results
Show Answer
**B** — The Extractor pattern's critical guardrails are: (1) "only include information explicitly stated in the source" — which prevents inference-based fabrication — and (2) a defined response for missing fields ("write Not specified") — which prevents the model from inventing information to fill empty fields. Without these two rules, extraction prompts frequently produce plausible-sounding but invented data, especially for fields that are contextually expected but absent from the actual source.Question 7
Elena builds a pattern called "Recommendation Builder" that combines the Analyzer and Generator patterns. What is this called, and what does the composite approach accomplish?
A) Pattern forking — creates two separate patterns from one B) Pattern composition — using the output of one pattern as input to another to produce a more complex result C) Pattern recursion — running the same pattern twice on the same content D) Pattern abstraction — removing detail from two patterns to create a simpler one
Show Answer
**B** — Pattern composition is the technique of chaining two or more patterns where the output of one becomes the input for the next. Elena's "Recommendation Builder" analyzes the situation (Analyzer), generates options (Generator), evaluates options against criteria (Analyzer again), and recommends (custom synthesis). This is more powerful than either pattern alone for her use case because consulting recommendations require both systematic analysis and creative option generation.Question 8
What is the primary maintenance failure that causes prompt libraries to become useless over time?
A) Storing patterns in too many locations B) Building patterns that are too specific to one task instance — they have no [bracket variables] and cannot be reused without significant rewriting C) Using too many patterns (library becomes too large to navigate) D) Not updating patterns when the AI platform updates its model
Show Answer
**B** — The most common maintenance failure is building prompts, not patterns: storing specific task instances without generalizing them into reusable templates. Without bracket variables, what looked like a reusable pattern is really just a saved prompt that only works for the original task. The second most common failure is not updating patterns after the task requirements evolve, leaving the library full of patterns that no longer produce good results.Question 9
Raj's "Code Reviewer" pattern is a customized version of Pattern 5 (The Critic/Reviewer). What makes it specific to his context, and why is this specificity valuable?
A) It specifies the programming language in the template, which improves code parsing B) It assigns a role with fintech-specific standards and specifies exactly which issue categories to review, making it produce consistent, security-focused code review every time C) It includes his company's code style guide as a fixed appendix D) It automatically runs the review three times and consolidates the results
Show Answer
**B** — Raj's pattern specifies "as a security-focused senior engineer at a fintech company" (role with domain context) and lists exactly four issue types to check: security vulnerabilities, error handling, test coverage gaps, and performance risks. This specificity means every code review covers the same ground, in the same order, against the standards most relevant to his context. The value is consistency: every PR reviewed with this pattern gets evaluated on the same criteria, not whatever happens to come to mind that day.Question 10
Pattern 11 (The Rewriter) asks you to specify both what to CHANGE and what to PRESERVE. Why is the "preserve" section as important as the "change" section?
A) The preserve section is optional — it's only needed for legal or compliance content B) Without explicit preservation instructions, the model may improve clarity by oversimplifying or omitting important content, solving one problem while creating another C) The preserve section prevents the model from making any changes beyond the specified ones, ensuring exact control D) The preserve section reduces response length, which makes the rewrite faster
Show Answer
**B** — The preserve section is critical because rewriting involves trade-offs. Simplifying language can lose technical precision. Shortening can cut important caveats. Converting technical terms to plain language can lose meaningful distinctions. By explicitly stating what must be preserved, you tell the model which trade-offs are not acceptable — giving it room to improve clarity while protecting the elements that must not change. Without this, "improve clarity" can easily mean "lose accuracy."Question 11
Why does Pattern 12 (The Brainstormer) specify diversity requirements like "at least 2 unconventional ideas" rather than just asking for creative results?
A) "Creativity" is not a recognized instruction that AI models respond to B) Without specific diversity constraints, models cluster around the most common solution type for the task, producing variations rather than genuinely different approaches C) Diversity requirements make the output shorter and easier to evaluate D) Legal constraints require AI to acknowledge when ideas are conventional vs. novel
Show Answer
**B** — Without diversity constraints, the Generator and Brainstormer patterns tend to produce variations on a single theme — the most statistically common approach to the problem type. "Generate 12 ideas for reducing churn" without diversity requirements will produce 12 variations of "improve onboarding" or "add a customer success team." Explicit diversity constraints ("at least 2 focused on the product, at least 2 on the relationship, range from quick wins to 6-month initiatives") force the model to cover different parts of the solution space.Question 12
What does it mean for a pattern to have "context override" as a failure mode?
A) The model ignores your role assignment when given long context B) When you paste very long content, the content can overwhelm the instructions, causing the model to follow the content's implicit structure rather than your specified format C) The pattern works for the original context but not for new contexts D) Adding too much context information makes the pattern too slow to use
Show Answer
**B** — Context override is a specific failure mode where long pasted content can "push" the instructions out of the model's effective attention. The model follows the patterns it sees in the content rather than the instructions at the top of the prompt. This is especially common for Extractor, Summarizer, and Rewriter patterns that process very long documents. The fix is to put key instructions at the end as well as the beginning, and to add explicit instructions like "follow the output format specified above exactly."Question 13
What is the "blank page problem" as it applies to AI prompting, and how do patterns address it?
A) The tendency of AI to produce blank outputs when given vague prompts — patterns reduce vagueness B) The cognitive effort required to construct a good prompt from scratch — patterns reduce initiation cost by providing a pre-built structure to fill in C) The lack of historical context when starting a new AI session — patterns provide that context D) The tendency to forget prompts that worked — patterns provide a memory system
Show Answer
**B** — The "blank page problem" in pattern context refers to cognitive load: the mental effort of constructing a good prompt from scratch. Research on knowledge work shows that templates reduce this initiation friction significantly — it is easier to fill in a structured template than to compose from nothing. Patterns solve this for AI prompting: instead of staring at a blank chat input and constructing the prompt from scratch, you open a template, fill in variables, and launch. The cognitive load reduction compounds across many tasks per day.Question 14
Pattern 8 (The Planner) recommends the "plan then execute" approach from Chapter 10. In the context of a pattern library, why is this approach particularly valuable?
A) It produces longer documents, which are more impressive to stakeholders B) It allows you to review and approve the plan structure before any content is generated, preventing the expensive rework of fixing structural problems after all content is written C) It reduces the number of patterns needed in your library D) It allows the AI to use different patterns for different sections automatically
Show Answer
**B** — The plan-then-execute approach is valuable in the Planner pattern because plans are structural decisions — getting the phases, milestones, and sequence wrong requires rewriting all the dependent content. By generating the plan structure first, reviewing it, and revising before executing, you ensure the content is developed within a framework you've validated. This is especially important when using AI for multi-section deliverables where sections build on each other.Question 15
Alex tracks that her AI copy revision rate dropped from 65% to 12% after building her five-example few-shot library and integrating it into her pattern. What does this demonstrate about the relationship between Chapter 10 techniques and Chapter 11 patterns?
A) Few-shot prompting is only effective when used as part of a pattern library B) Patterns work independently of techniques — the improvement came from the pattern structure, not the few-shot examples C) Advanced techniques from Chapter 10 can be embedded directly into patterns, making those techniques systematic and consistently applied rather than ad hoc D) The revision rate improvement was coincidental — the pattern structure alone would have produced the same result