AI for Everyone study guide · Lesson 1 of 14

A Good Prompt Has Parts

Name the Task · Give Relevant Context · Set Constraints and Format · Define Success

This is the lesson's study guide from the FlashFeed app, plus its key facts and a self-test.

What you'll be able to do

  • Build a prompt with a clear task, relevant context, useful constraints, and a testable output.

Name the Task

A useful prompt begins with a verb and an outcome. “Help with my trip” leaves the system to guess whether you want ideas, an itinerary, a budget, or packing advice. “Draft a two-day walking itinerary” names an observable deliverable. The wording does not need to be formal; it needs to make the job visible.

Before sending, underline the action and circle the deliverable. If either is missing, add it. For an unfamiliar task, ask the system to restate the assignment and list open questions before it answers. That small check exposes mismatched assumptions while they are still cheap to fix.

Three ideas to retain:

  • Start with an Action: A strong prompt states what the system should do: draft, compare, classify, extract, critique, plan, or explain. A verb makes the requested behavior easier to evaluate.
  • Name the Deliverable: The same topic can produce a memo, table, checklist, lesson, or set of questions. Naming the deliverable prevents the system from choosing the container for you.
  • Separate Goal from Topic: “Retirement” is a topic. “Explain three retirement-account differences to a new employee” is a goal. A goal combines subject, action, and intended result.

Key term: Task verb means “the action the system should perform.”

Give Relevant Context

Context is information that would change a competent answer. Audience, purpose, setting, prior decisions, and source material often matter. A beginner needs different explanations than a specialist; a private brainstorming note needs a different tone than a public announcement.

More context is not automatically better. Long, unfiltered dumps can bury the decisive detail, and models do not use every position in a long context equally well. Provide the smallest relevant packet, label its sections, and distinguish facts from preferences. If something is unknown, say so instead of inviting the system to invent it.

Three ideas to retain:

  • Context Changes the Answer: Audience, purpose, constraints, and source material determine what a useful answer looks like. The same task can need different content in a classroom, clinic, or kitchen.
  • Relevant Beats Exhaustive: Context should earn its place by changing the output. A focused brief is usually easier to use and verify than a long dump of unrelated history.
  • Label Inputs Clearly: Headings such as BACKGROUND, SOURCE TEXT, and REQUIREMENTS help separate evidence from instructions and make complex prompts easier to inspect.

Key term: Audience context means “who will use or read the output.”

Set Constraints and Format

Constraints turn taste into instructions. Length, tone, reading level, required sections, excluded content, deadline, and output format are all legitimate boundaries. The best constraints reflect the real use of the result: a table for comparison, bullets for scanning, or JSON for software.

Constraints can conflict. “Comprehensive,” “one paragraph,” and “include every exception” may not fit together. Rank the requirements or state which can flex. Prefer positive directions—what to include and how to organize it—while using exclusions for genuine hazards or scope boundaries.

Three ideas to retain:

  • Format Is Part of the Task: A correct answer can still be unusable if it arrives in the wrong structure. Specify a table, numbered plan, email, schema, or other container when the container matters.
  • Constraints Need Priorities: Length, completeness, tone, and detail can pull in different directions. Ranking constraints tells the system what should win when all cannot be satisfied.
  • Positive Directions Are Clearer: “Use three short sections with concrete examples” provides a target. A long list of prohibitions may define only what the output should not be.

Key term: Hard constraint means “a boundary the output must satisfy.”

Define Success

Acceptance criteria describe what must be true of a satisfactory answer. They might require every recommendation to cite supplied evidence, every recipe to avoid an allergen, or every row to contain a date. Criteria make revision less personal: instead of “make it better,” you can point to the missed requirement.

A miniature example often communicates style faster than adjectives. Show one short input and the kind of output you want, then explain which features should carry over. End by asking the model to check its draft against the criteria and flag any item it could not meet. That check is evidence, not a guarantee.

Three ideas to retain:

  • Criteria Make Quality Testable: Acceptance criteria translate “good” into observable conditions such as required evidence, fields, length, coverage, or prohibited content.
  • Examples Carry Tacit Structure: A small example can reveal naming, granularity, ordering, and tone that would take many abstract instructions to describe.
  • Self-Checks Still Need Review: A model can compare its answer with a checklist, but it can also miss its own error. Treat the check as a useful pass, not independent verification.

Key term: Acceptance criterion means “a condition the result must meet.”

Sources

Key terms

Task specification
A concise statement of the action, subject, and intended result.
Context
Information supplied with a request that helps determine a relevant response.
Output format
The structure or representation in which a response should be delivered.
Rubric
A set of criteria used to judge the quality of an output.

Key facts

Name the Deliverable
Name the DeliverableAI-generated (gpt-image-1.5)

Name the Deliverable

The same topic can produce a memo, table, checklist, lesson, or set of questions. Naming the deliverable prevents the system from choosing the container for you.

Separate Goal from Topic

“Retirement” is a topic. “Explain three retirement-account differences to a new employee” is a goal. A goal combines subject, action, and intended result.

Relevant Beats Exhaustive
Relevant Beats ExhaustiveAI-generated (gpt-image-1.5)

Relevant Beats Exhaustive

Context should earn its place by changing the output. A focused brief is usually easier to use and verify than a long dump of unrelated history.

Label Inputs Clearly

Headings such as BACKGROUND, SOURCE TEXT, and REQUIREMENTS help separate evidence from instructions and make complex prompts easier to inspect.

Constraints Need Priorities

Length, completeness, tone, and detail can pull in different directions. Ranking constraints tells the system what should win when all cannot be satisfied.

Positive Directions Are Clearer

“Use three short sections with concrete examples” provides a target. A long list of prohibitions may define only what the output should not be.

Examples Carry Tacit Structure

A small example can reveal naming, granularity, ordering, and tone that would take many abstract instructions to describe.

Self-Checks Still Need Review
Self-Checks Still Need ReviewAI-generated (gpt-image-1.5)

Self-Checks Still Need Review

A model can compare its answer with a checklist, but it can also miss its own error. Treat the check as a useful pass, not independent verification.

Matched pairs

Each item on the left goes with the one on the right.

Task verb
The action the system should perform
Deliverable
The concrete output you expect
Audience context
Who will use or read the output
Situational context
The circumstances that shape the task
Hard constraint
A boundary the output must satisfy
Soft preference
A desirable feature that may flex
Acceptance criterion
A condition the result must meet
Reference example
A sample that demonstrates desired features

Self-test

Answer each question from memory, then check your work against the answer key.

  1. 1. Which opening is most actionable?

    1. ATell me about plans
    2. BWhat do you think?
    3. CCompare these three plans
    4. DPlans are interesting
  1. 2. What is missing from “Help with my résumé”?

    1. AA creative persona
    2. BA specific outcome
    3. CA model nickname
    4. DA longer greeting
  2. 3. Which detail most changes a safety handout?

    1. AThe font is your favorite
    2. BThe prompt says please
    3. CThe model has a logo
    4. DThe readers are children
  3. 4. What is the best context rule?

    1. AInclude details that change the answer
    2. BHide uncertainty
    3. CRepeat the task five times
    4. DPaste every related document
  4. 5. Which instruction defines a format?

    1. ABe intelligent
    2. BThink carefully
    3. CUse your knowledge
    4. DReturn a three-column table

Go deeper

Answer key

  1. 1. C Compare these three plans“Compare” names a visible operation and identifies the supplied choices.
  2. 2. B A specific outcomeThe request does not say whether to draft, edit, tailor, or critique.
  3. 3. D The readers are childrenAge changes vocabulary, examples, and safety framing.
  4. 4. A Include details that change the answerUseful context is selected for relevance, not sheer volume.
  5. 5. D Return a three-column tableA table with named columns is an observable structure.

Image credits

  • Name the Deliverable: AI-generated (gpt-image-1.5)
  • Relevant Beats Exhaustive: AI-generated (gpt-image-1.5)
  • Self-Checks Still Need Review: AI-generated (gpt-image-1.5)

Download all 14 lessons: PDF · EPUB