Write Better AI Prompts, Faster
This prompt engineering cheat sheet gives you practical formulas, frameworks, techniques, modifiers, and ready-to-use prompt templates for producing clearer and more reliable AI responses.
Use it with ChatGPT, Claude, Gemini, Copilot, and other generative AI tools for writing, coding, research, learning, business, and creative work.
- Prompt formulas
- Copy-ready templates
- Practical examples
- Advanced techniques
Role + Task + Context + Constraints + Output Format
Search This Cheat Sheet
Find prompt formulas, frameworks, techniques, modifiers, and templates.
Matching sections will remain visible while unrelated sections are hidden.
Prompt Engineering Quick Reference
Strong prompts clearly define what the AI should do, why it should do it, and how the final response should be presented.
Role + Task + Context + Constraints + Output Format + Examples
Role
Define the expertise, perspective, or professional role the AI should adopt.
Act as a senior UX designer.
Task
State the exact action you want the AI to complete using a clear instruction verb.
Create a usability checklist.
Context
Provide the background information the AI needs to understand the situation.
The checklist is for a mobile banking app.
Constraints
Specify requirements, boundaries, exclusions, length, tone, or other limitations.
Use plain English and limit it to 10 items.
Output Format
Describe how the response should be structured and presented.
Return a Markdown table with three columns.
Examples
Include one or more examples when the desired style or pattern could be misunderstood.
Follow the format shown in this example: …
Universal Prompt Template
Replace the bracketed placeholders with information about your task. Remove any line that is not relevant.
Build Your Prompt
Act as [role or area of expertise].
Your task is to [specific action].
Context:
[Provide the background information, goal, audience, and any relevant data.]
Requirements:
- [Requirement or constraint 1]
- [Requirement or constraint 2]
- [Requirement or constraint 3]
Output format:
[Describe the required structure, length, tone, and formatting.]
Example or reference:
[Include an example when useful.]
If essential information is missing, ask concise clarifying questions before completing the task.UX Audit Prompt
Act as a senior UX researcher.
Your task is to create a usability audit checklist.
Context:
The checklist will be used to evaluate the onboarding flow of a mobile banking app designed for first-time users.
Requirements:
- Use plain English.
- Include no more than 10 checklist items.
- Focus on clarity, accessibility, trust, and error prevention.
- Make every item specific and actionable.
Output format:
Return a Markdown table with the columns Check, Why It Matters, and Priority.
Use a professional and concise tone.Anatomy of an Effective Prompt
An effective prompt gives the AI enough direction to complete the task without adding unnecessary instructions. Each part below has a specific purpose.
SEO Content Brief Prompt
Act as an experienced SEO content strategist.
Create a detailed content brief for an article about home solar panels.
The article is for first-time homeowners comparing renewable energy options. Its goal is to educate readers and support an informed purchasing decision.
Use plain English, avoid exaggerated claims, and do not invent statistics. Include important questions a homeowner should ask an installer.
Return the brief in Markdown with a proposed title, search intent, audience, primary sections, FAQs, and internal-link opportunities.
The brief should be practical enough for a freelance writer to use without additional instructions.Role
Establishes the expertise and perspective the AI should use.
Task
Defines the primary action and the deliverable you expect.
Context
Explains the audience, situation, purpose, and background.
Constraints
Sets boundaries for accuracy, tone, scope, length, and exclusions.
Output Format
Specifies how the final answer should be organized and presented.
Success Criteria
Describes what a useful, complete, and high-quality result looks like.
Basic Prompt Formulas
Start with the shortest formula that can communicate your goal. Add context, constraints, and examples only when they improve the result.
Direct Instruction
Action + Subject + Output Format
Summarize the following article in five bullet points for a general audience.
Expert Role
Role + Task + Context + Requirements
Act as a cybersecurity trainer. Explain phishing to new employees using plain English, one realistic example, and a five-item prevention checklist.
Content Transformation
Transform + Input + Target + Preserve
Rewrite the text below as a concise customer email. Preserve every date, price, and product name. Use a friendly but professional tone.
Analysis and Comparison
Analyze + Options + Criteria + Recommendation
Compare the two project plans using cost, implementation time, maintenance, and risk. Return a table followed by a recommendation with stated assumptions.
Review and Improve
Review + Criteria + Weaknesses + Revised Version
Review the draft below for clarity, accuracy, structure, and tone. Identify the three most important weaknesses, explain each briefly, and then provide an improved version.
Prompting Techniques
Different tasks require different prompting strategies. Start with the simplest technique and add examples, stages, or evaluation only when they improve the result.
Zero-Shot Prompting
Ask the AI to complete a task without providing a demonstration. The model relies on its existing knowledge and your instructions.
Explain compound interest to a high-school student. Use plain English, one everyday example, and no more than 150 words.Few-Shot Prompting
Provide a small number of input-and-output examples so the AI can follow the desired pattern, style, or classification rule.
Classify each review as Positive, Neutral, or Negative.
Review: "Excellent quality and fast delivery."
Label: Positive
Review: "The product is acceptable but nothing special."
Label: Neutral
Review: "The item stopped working after two days."
Label: Negative
Review: "It works well, although delivery was late."
Label:Task Decomposition
Divide a complex assignment into smaller stages with a clear deliverable for each stage.
Help me create a launch plan for an online course.
Divide the task into these stages:
1. Define the target audience and value proposition.
2. Create the course outline.
3. Plan the launch content.
4. Build a four-week promotion schedule.
5. Define success metrics.
Complete one stage at a time. Present each stage as a concise, actionable deliverable.Iteration, Evaluation, and Grounding
Use these techniques when a first response needs refinement, quality control, or stronger factual grounding.
Iterative Prompting
Improve the result through focused follow-up prompts instead of attempting to perfect every requirement in one instruction.
Revise the draft using the following feedback:
- Make the opening more specific.
- Remove repeated ideas.
- Add one practical example.
- Keep the final version under 500 words.
- Preserve all verified facts.
Return only the revised draft.Self-Critique and Revision
Ask the AI to evaluate a draft against explicit criteria and then produce an improved version.
Evaluate the draft below for clarity, accuracy, completeness, tone, and unnecessary repetition.
First, provide a concise list of specific issues. Then produce a revised version that resolves those issues without adding unsupported claims.
Draft:
[Paste draft]Retrieval-Grounded Prompting
Require the response to rely on supplied documents, excerpts, or verified source material rather than unsupported assumptions.
Answer the question using only the provided sources.
For every factual claim, cite the relevant source as [Source 1], [Source 2], and so on. If the sources do not contain enough information, state what is missing instead of guessing.
Question:
[Enter question]
Sources:
[Paste sources]Prompt Frameworks
Prompt frameworks provide a repeatable structure for complex instructions. Their names and letter meanings sometimes vary, so use them as practical checklists rather than rigid rules.
CREATE Framework
A flexible framework for building detailed prompts with clear instructions, examples, adjustments, and output requirements.
Character — Define the role or expertise.
Request — State the task or desired action.
Examples — Show the desired pattern when useful.
Adjustments — Set tone, limits, and requirements.
Type of Output — Specify the response format.
Extras — Add final checks or special instructions.
Content creation, business tasks, planning, analysis, and detailed one-off prompts.
CREATE Prompt
Character:
Act as [role or area of expertise].
Request:
[Describe the specific task and goal.]
Examples:
[Provide an example or reference when useful.]
Adjustments:
[Define tone, audience, length, constraints, and exclusions.]
Type of output:
[Specify the required format and structure.]
Extras:
[Add quality checks, priorities, or instructions for missing information.]CO-STAR Framework
A detailed framework for controlling the purpose, writing style, tone, audience, and final response.
Context — Explain the background and situation.
Objective — Define the goal and desired outcome.
Style — Choose the writing or presentation style.
Tone — Set the emotional quality of the response.
Audience — Identify who will use or read the output.
Response — Specify the format and final deliverable.
Marketing, email, social media, presentations, educational content, and audience-specific communication.
CO-STAR Prompt
Context:
[Describe the background, situation, and relevant information.]
Objective:
[State the specific goal or result.]
Style:
[Define the writing or presentation style.]
Tone:
[Choose the desired tone.]
Audience:
[Describe the intended reader or user.]
Response:
[Specify the format, structure, length, and required elements.]RACE Framework
A compact framework for assigning a role, defining the action, supplying context, and describing the expected result.
Role — Define the expertise or perspective.
Action — State exactly what the AI should do.
Context — Add the relevant background and input.
Expectations — Describe the required result and quality.
Everyday business tasks, analysis, planning, problem-solving, and concise professional prompts.
RACE Prompt
Role:
Act as [role or area of expertise].
Action:
[Describe the specific task.]
Context:
[Provide the background, audience, goal, and relevant input.]
Expectations:
[Define the required format, level of detail, constraints, and success criteria.]RISEN Framework
A structured framework for complex assignments that need explicit instructions, defined stages, a clear end goal, and firm boundaries.
Role — Assign the relevant expertise or perspective.
Instructions — Describe the primary task clearly.
Steps — Define the stages or process to follow.
End Goal — State the final outcome and purpose.
Narrowing — Set scope, limits, exclusions, and priorities.
Project planning, research, strategy, coding workflows, detailed analysis, and multi-stage assignments.
RISEN Prompt
Role:
Act as [role or area of expertise].
Instructions:
[Describe the main task and important requirements.]
Steps:
1. [First stage]
2. [Second stage]
3. [Third stage]
End goal:
[Define the final outcome and how it will be used.]
Narrowing:
[Set the scope, constraints, exclusions, priorities, and output format.]CARE Framework
A straightforward framework for describing a situation, requesting an action, defining the intended result, and showing a useful example.
Context — Describe the situation and relevant background.
Action — State the task the AI should complete.
Result — Define the desired outcome and format.
Example — Provide a model, reference, or desired pattern.
Customer service, workplace communication, content rewriting, training material, and practical everyday requests.
CARE Prompt
Context:
[Describe the situation, audience, and relevant background.]
Action:
[State exactly what the AI should create, analyze, or improve.]
Result:
[Define the desired outcome, format, length, and tone.]
Example:
[Provide an example or reference that demonstrates the preferred result.]Prompt Modifiers
Prompt modifiers are short instructions that control how an AI response should sound, who it is for, how detailed it should be, and how it should be presented.
For a searchable library of copy-ready instructions for tone, style, length, format, audience, accuracy, and more, explore the AI Prompt Modifiers Cheat Sheet .
Tone
Controls the emotional quality and attitude of the response.
Use a professional, reassuring, and concise tone.
Audience
Adjusts vocabulary, explanations, assumptions, and examples for the intended reader.
Write for beginners with no prior technical knowledge.
Length
Sets an approximate or exact size for the final response.
Limit the response to five bullet points and 150 words.
Level of Detail
Defines how deeply the AI should explain the subject.
Explain the process step by step and include one practical example for each stage.
Format, Style, Perspective, and Evidence
Use these modifiers to control the structure, presentation, viewpoint, and factual grounding of the response.
Format
Defines the structure in which the final response should be returned.
Return the answer as a Markdown table with the columns Option, Benefits, Risks, and Best Use Case.
Style
Controls the writing approach, presentation method, or creative style.
Write in a clear technical-documentation style with short sections, descriptive headings, and practical examples.
Perspective
Sets the viewpoint, professional lens, or position used in the answer.
Evaluate the proposal from both a customer perspective and an operations-manager perspective.
Evidence and Sources
Sets requirements for factual support, uncertainty, and source use.
Use only the supplied sources. Cite each factual claim and clearly state when the available evidence is insufficient.
Prompt Modifier Builder
Select the controls you need. The builder combines them into a copy-ready instruction that can be added to any prompt.
Add This to Your Prompt
Select one or more modifiers to build your instruction.Tip: use only the modifiers that materially improve the result.
AI Output Formats
Specifying the output format makes responses easier to read, reuse, compare, or pass into another tool. Describe both the structure and the required fields.
Bullet List
Best for summaries, recommendations, ideas, and key takeaways.
Return the answer as five concise bullet points. Begin each bullet with a bold action verb and keep each item under 25 words.
- **Identify** the primary goal.
- **Collect** the required information.
- **Compare** the available options.
- **Select** the strongest approach.
- **Review** the result before publishing.Markdown Table
Best for comparisons, specifications, plans, and categorized data.
Return the answer as a Markdown table with the columns Option, Key Benefit, Limitation, Cost Level, and Best Use Case.
| Option | Key Benefit | Limitation | Cost Level | Best Use Case |
|--------|-------------|------------|------------|---------------|
| A | ... | ... | ... | ... |JSON
Best for APIs, automation, applications, and machine-readable data.
Return only valid JSON. Use the properties title, summary, priority, tags, and next_actions. Do not include Markdown fences or explanatory text.
{
"title": "...",
"summary": "...",
"priority": "high",
"tags": ["...", "..."],
"next_actions": ["...", "..."]
}Checklists, Guides, and Reports
These formats work well when the response needs to support action, explain a process, or present a formal analysis.
Checklist
Best for reviews, preparation, quality control, and repeatable tasks.
Return the answer as an actionable checklist. Group related items under descriptive headings and begin every item with an unchecked Markdown checkbox.
## Preparation
- [ ] Define the objective
- [ ] Collect required information
## Review
- [ ] Verify accuracy
- [ ] Confirm all requirementsStep-by-Step Guide
Best for tutorials, workflows, setup instructions, and processes.
Explain the process as a numbered step-by-step guide. For every step, include the action, why it matters, and the expected result.
1. **Action**
- What to do: ...
- Why it matters: ...
- Expected result: ...
2. **Next action**
- What to do: ...
- Why it matters: ...
- Expected result: ...Structured Report
Best for research, audits, evaluations, and business analysis.
Return a structured report with an Executive Summary, Background, Key Findings, Evidence, Risks, Recommendations, and Next Steps. Clearly separate facts, assumptions, and uncertainties.
# Report Title
## Executive Summary
...
## Key Findings
...
## Risks
...
## Recommendations
...
## Next Steps
...Ready-to-Use Prompt Templates
Replace the bracketed placeholders with your own information. Keep only the instructions that are relevant to your task.
Summarize Content
Extract the central argument, evidence, conclusions, and practical implications from supplied material.
Summarize the content below for [target audience].
Include:
- The central argument or purpose
- The most important supporting points
- Key evidence, examples, or data
- The main conclusion
- Practical implications or next actions
Requirements:
- Use plain English.
- Preserve important qualifications and uncertainty.
- Do not add claims that are not present in the source.
- Keep the summary under [length].
Output format:
[Bullet list / executive summary / structured notes]
Content:
[Paste content]Brainstorm Ideas
Generate varied ideas while keeping them aligned with a defined goal, audience, and set of constraints.
Generate [number] distinct ideas for [goal or project].
Context:
[Describe the situation, audience, available resources, and desired outcome.]
Constraints:
- [Constraint 1]
- [Constraint 2]
- [Constraint 3]
Make the ideas meaningfully different rather than minor variations.
For each idea, provide:
1. A short name
2. A one-sentence concept
3. The primary benefit
4. A possible limitation
5. The first practical stepCompare Options and Create a Plan
Use these templates to evaluate alternatives and convert a goal into a structured sequence of actions.
Compare Options
Compare alternatives using consistent criteria, trade-offs, evidence, and a reasoned recommendation.
Compare the following options for [decision or goal]:
Options:
- [Option 1]
- [Option 2]
- [Option 3]
Evaluation criteria:
- [Criterion 1]
- [Criterion 2]
- [Criterion 3]
- [Criterion 4]
For each option, explain:
1. Main benefits
2. Important limitations
3. Cost or resource requirements
4. Relevant risks
5. Best use case
Return a Markdown comparison table followed by a balanced recommendation.
Clearly state any assumptions or missing information.Create an Action Plan
Turn a defined objective into prioritized stages, concrete tasks, responsibilities, and measurable outcomes.
Create a practical action plan for achieving this goal:
Goal:
[Describe the desired outcome.]
Current situation:
[Explain the starting point, available resources, and relevant constraints.]
Deadline:
[Enter deadline or time frame.]
Divide the plan into logical stages.
For each stage, include:
- Objective
- Specific tasks
- Priority
- Required resources
- Responsible person or role
- Target date
- Success measure
- Possible risk and mitigation
Finish with the three actions that should be completed first.Writing and Content Prompts
Use AI to plan, draft, edit, and repurpose content while preserving accuracy, audience relevance, and a consistent voice.
Create an Article Outline
Build a logical article structure based on topic, audience, search intent, and reader questions.
Act as an experienced content strategist.
Create a detailed article outline for:
Topic:
[Enter topic]
Target audience:
[Describe the reader]
Primary goal:
[Educate / compare / persuade / explain / help complete a task]
Search intent:
[Informational / commercial / transactional / navigational]
Requirements:
- Propose a clear working title.
- Organize the outline with logical H2 and H3 headings.
- State the purpose of every section.
- Include important questions the reader expects answered.
- Identify places for examples, tables, checklists, or visuals.
- Avoid overlapping sections.
- Add a concise FAQ outline.
Return the result in Markdown.Write a First Draft
Turn an approved outline and source material into a clear, useful, and well-structured first draft.
Write a first draft using the outline and source material below.
Audience:
[Describe the reader]
Purpose:
[Describe what the reader should understand or do]
Tone and style:
[Define tone, voice, and level of detail]
Requirements:
- Follow the outline in order.
- Use descriptive headings and short paragraphs.
- Explain unfamiliar terms.
- Include practical examples where useful.
- Preserve important qualifications and uncertainty.
- Do not invent facts, statistics, quotes, or sources.
- Mark any section that lacks sufficient source information with [SOURCE NEEDED].
Outline:
[Paste outline]
Verified source material:
[Paste source material]Improve and Repurpose Content
Strengthen an existing draft or adapt one verified source into platform-specific content without changing its meaning.
Edit and Improve Content
Improve clarity, structure, flow, tone, and readability while preserving the original meaning and verified facts.
Act as a professional editor.
Review and improve the draft below for:
- Clarity
- Logical structure
- Readability
- Concision
- Tone consistency
- Repetition
- Grammar and punctuation
Requirements:
- Preserve the original meaning.
- Preserve every verified fact and qualification.
- Do not add unsupported claims.
- Replace vague wording with specific language where the source allows it.
- Use short paragraphs and descriptive headings.
- Keep the voice natural rather than overly promotional.
First, list the five most important issues.
Then return the fully revised draft.
Draft:
[Paste draft]Repurpose Content
Convert one source into several platform-specific assets while retaining its core message and factual accuracy.
Repurpose the source content below into:
1. A five-post social media series
2. A short email newsletter
3. A 60-second video script
4. A concise LinkedIn post
5. Three headline variations
Target audience:
[Describe the audience]
Brand voice:
[Describe the tone and style]
Requirements:
- Adapt each asset to its platform.
- Preserve the source's core message and verified facts.
- Do not repeat identical wording across formats.
- Do not invent statistics, quotations, or examples.
- Include a relevant call to action for each format.
Source content:
[Paste content]Business and Productivity Prompts
Use these templates to improve workplace communication, document decisions, organize meetings, and turn recurring processes into clear instructions.
Write a Professional Email
Create a clear email with the right context, tone, request, and next action.
Write a professional email using the information below.
Recipient:
[Who will receive the email?]
Relationship:
[Colleague / customer / manager / supplier / other]
Purpose:
[Explain why you are writing.]
Important context:
[Provide the necessary background.]
Requested action:
[State what the recipient should do.]
Deadline:
[Enter a deadline if relevant.]
Tone:
[Professional / friendly / diplomatic / firm / appreciative]
Requirements:
- Write a clear subject line.
- State the purpose early.
- Keep the email concise.
- Use specific rather than vague language.
- End with a clear next step.
- Do not invent names, dates, commitments, or facts.Summarize a Meeting
Convert meeting notes or a transcript into decisions, action items, owners, deadlines, and unresolved questions.
Summarize the meeting notes or transcript below.
Return the result with these sections:
1. Meeting Purpose
2. Key Discussion Points
3. Decisions Made
4. Action Items
5. Owners
6. Deadlines
7. Open Questions
8. Risks or Blockers
9. Recommended Follow-Up
Requirements:
- Use concise, neutral language.
- Do not treat suggestions as confirmed decisions.
- Do not invent owners or deadlines.
- Mark missing owners or dates as "Not assigned."
- Present action items in a Markdown table.
Meeting content:
[Paste notes or transcript]Document Decisions and Workflows
Turn analysis into a concise recommendation or convert a repeatable activity into a clear standard operating procedure.
Create a Decision Memo
Present a decision, alternatives, evidence, risks, and a recommendation in a concise business format.
Create a concise decision memo using the information below.
Decision:
[Describe the decision that must be made.]
Goal:
[Define the desired outcome.]
Options:
- [Option 1]
- [Option 2]
- [Option 3]
Evaluation criteria:
- [Criterion 1]
- [Criterion 2]
- [Criterion 3]
Evidence:
[Paste verified facts, data, or analysis.]
Return the memo with:
1. Executive Summary
2. Decision Required
3. Background
4. Options Considered
5. Benefits and Risks
6. Recommendation
7. Assumptions and Uncertainties
8. Immediate Next Steps
Do not invent evidence. Clearly distinguish facts from assumptions.Build a Standard Operating Procedure
Convert process knowledge into a repeatable SOP with roles, steps, quality checks, and exception handling.
Create a Standard Operating Procedure for:
Process:
[Name and describe the process.]
Purpose:
[Explain the desired outcome.]
People or roles involved:
[List responsible roles.]
Required tools or inputs:
[List systems, documents, or materials.]
Known process steps:
[Paste existing notes or instructions.]
Return the SOP with:
1. Purpose
2. Scope
3. Roles and Responsibilities
4. Prerequisites
5. Numbered Procedure
6. Quality-Control Checks
7. Exceptions and Escalation
8. Required Records
9. Review Schedule
For every procedure step, include the action, responsible role, and expected result.
Mark missing information as [TO CONFIRM] rather than guessing.Coding Prompts
Better coding prompts define the environment, requirements, constraints, expected behavior, and verification method instead of requesting code without context.
Generate Code
Request production-ready code with explicit requirements, environment details, constraints, and verification steps.
Act as a senior [language or framework] developer.
Build:
[Describe the feature or component.]
Environment:
- Language and version: [Enter version]
- Framework and version: [Enter version]
- Runtime or platform: [Enter environment]
- Existing dependencies: [List dependencies]
Requirements:
- [Functional requirement 1]
- [Functional requirement 2]
- [Functional requirement 3]
Constraints:
- [Performance, accessibility, security, compatibility, or dependency limits]
Expected input:
[Describe input]
Expected output:
[Describe output]
Return:
1. Complete code
2. Concise implementation explanation
3. Setup or integration instructions
4. Important edge cases
5. Tests or verification steps
If essential technical information is missing, ask focused clarification questions before writing the code.Debug an Error
Diagnose a reproducible problem using the code, error message, environment, expected behavior, and attempted fixes.
Act as an experienced debugging specialist.
Diagnose the following problem.
Environment:
[Language, framework, versions, operating system, browser, or runtime]
Expected behavior:
[Describe what should happen.]
Actual behavior:
[Describe what happens instead.]
Exact error message:
[Paste the complete error.]
Relevant code:
[Paste the smallest reproducible code sample.]
Steps to reproduce:
1. [Step 1]
2. [Step 2]
3. [Step 3]
Already attempted:
[List attempted fixes and results.]
Return:
1. Most likely root cause
2. Evidence supporting the diagnosis
3. Minimal safe fix
4. Corrected code
5. How to verify the fix
6. Any remaining risks or assumptions
Do not recommend unrelated rewrites unless they are necessary to solve the problem.Refactor and Test Code
Improve maintainability without changing behavior, then verify the result with focused tests and edge-case coverage.
Refactor Existing Code
Improve readability, maintainability, or performance while preserving documented behavior and public interfaces.
Act as a senior software engineer.
Refactor the code below to improve:
- Readability
- Maintainability
- Naming
- Duplication
- Error handling
- Performance where measurable
Constraints:
- Preserve all documented behavior.
- Preserve public function signatures and output formats.
- Do not add new dependencies.
- Do not change unrelated code.
- Follow the existing project style.
- Explain any assumption that could affect behavior.
Return:
1. A concise list of identified problems
2. The refactored code
3. A summary of each meaningful change
4. Potential compatibility risks
5. Tests that should be run
Existing code:
[Paste code]Create Tests and Review Edge Cases
Generate focused tests for expected behavior, failures, boundaries, invalid inputs, and important regressions.
Act as a software test engineer.
Create tests for the code below.
Environment:
- Language and version: [Enter version]
- Test framework: [Enter framework]
- Runtime: [Enter environment]
Expected behavior:
[Describe the requirements.]
Include tests for:
- Normal successful inputs
- Boundary values
- Empty and missing inputs
- Invalid types or formats
- Expected failures
- Important regressions
- Security-sensitive cases where relevant
Return:
1. A test-case table with scenario, input, expected result, and priority
2. Complete test code
3. Required setup or fixtures
4. Uncovered risks or assumptions
5. Commands for running the tests
Code:
[Paste code]Learning and Research Prompts
Use AI to explain difficult subjects, organize learning, compare evidence, and synthesize supplied research without hiding uncertainty or inventing sources.
Explain a Concept
Build understanding progressively with a definition, analogy, example, common mistake, and knowledge check.
Act as an experienced teacher.
Explain the following concept:
Concept:
[Enter concept]
Learner level:
[Beginner / intermediate / advanced]
Existing knowledge:
[Describe what the learner already understands.]
Learning goal:
[Describe what the learner should be able to explain or do afterward.]
Structure the explanation as:
1. One-sentence definition
2. Plain-English explanation
3. Helpful analogy
4. Practical example
5. Step-by-step breakdown
6. Common misconception
7. When the concept is useful
8. Three-question knowledge check
Define unfamiliar terms and clearly label any simplifications or uncertainty.Create a Study Plan
Convert a learning goal, deadline, schedule, and current ability into a practical study roadmap.
Create a realistic study plan for:
Subject or skill:
[Enter subject]
Current level:
[Describe current knowledge.]
Goal:
[Describe the desired ability or exam result.]
Time available:
[Hours per day or week]
Deadline:
[Enter date or number of weeks]
Preferred learning methods:
[Reading / video / exercises / projects / discussion]
Return:
1. A week-by-week roadmap
2. Topics in prerequisite order
3. Practice activities
4. Active-recall questions
5. Review and revision sessions
6. Progress checkpoints
7. A final assessment or project
Keep the workload within the available time and identify any goal that may be unrealistic.Synthesize and Compare Evidence
Use supplied sources to identify agreements, disagreements, evidence quality, limitations, and unanswered questions.
Synthesize Research Sources
Combine several supplied sources into a structured overview without treating every claim as equally reliable.
Synthesize the supplied sources to answer this research question:
Research question:
[Enter question]
Use only the provided sources.
Return:
1. Executive Summary
2. Major Themes
3. Areas of Agreement
4. Areas of Disagreement
5. Strongest Supporting Evidence
6. Important Limitations
7. Unanswered Questions
8. Practical Implications
Requirements:
- Cite claims as [Source 1], [Source 2], and so on.
- Distinguish source findings from your synthesis.
- Do not invent citations or missing information.
- Note differences in definitions, methods, samples, or dates.
- State when the evidence is insufficient for a conclusion.
Sources:
[Paste and label each source]Compare Conflicting Sources
Examine why sources reach different conclusions and identify which claims are best supported.
Compare the conflicting sources below.
Question or disputed claim:
[Enter question]
For each source, identify:
- Main conclusion
- Evidence used
- Method or reasoning
- Important assumptions
- Limitations
- Publication date and relevance
Then explain:
1. Where the sources agree
2. Where they disagree
3. Whether they use different definitions or measurements
4. Which claims have the strongest support
5. What additional evidence would resolve the disagreement
Return a comparison table followed by a cautious conclusion.
Use only the supplied material and cite every source-specific claim.
Sources:
[Paste and label each source]Image-Generation Prompts
Effective image prompts describe the subject, scene, composition, lighting, color palette, visual style, and required constraints in a clear order.
Subject + Action + Environment + Composition + Lighting + Color + Style + Constraints
Photorealistic Image
Describe a realistic subject, environment, composition, lighting conditions, mood, and photographic treatment.
Create a photorealistic image of [main subject] [action or pose].
Environment:
[Describe the location, background, weather, time of day, and relevant objects.]
Composition:
[Close-up / medium shot / wide shot], viewed from [camera angle]. Place the subject [position in frame] with [foreground and background relationship].
Lighting:
[Soft natural light / dramatic side light / studio lighting / golden-hour light].
Color palette:
[Describe dominant and accent colors.]
Mood:
[Describe the intended atmosphere.]
Photographic treatment:
[Editorial / lifestyle / product / documentary / cinematic photography], realistic textures, natural proportions, and controlled depth of field.
Constraints:
[Aspect ratio, orientation, empty space for text, elements to include, and elements to avoid.]Flat Vector Illustration
Create a clean, scalable-looking illustration with simple forms, controlled colors, and a clear visual hierarchy.
Create a clean flat vector-style illustration of [subject or concept].
Visual concept:
[Describe the action, relationship, or idea the image should communicate.]
Composition:
Use a clear focal point, balanced spacing, simple geometric forms, and an uncluttered background.
Color palette:
Use [primary color], [secondary color], and [accent color] with accessible contrast.
Style:
Modern editorial illustration, crisp outlines, minimal shading, consistent stroke weight, and no photorealistic textures.
Requirements:
- Keep small details readable.
- Use a limited color palette.
- Leave [location] available for headline text.
- Do not include logos, watermarks, or unintended text.
- Format: [orientation and aspect ratio].Product Images and Infographic Concepts
Create a controlled product scene or generate a visual-layout concept for information that will be finalized in a design tool.
Product Image
Define the product, surface, environment, composition, lighting, branding constraints, and required space.
Create a premium product image featuring [product].
Product details:
[Describe shape, material, color, finish, packaging, and important visible features.]
Scene:
Place the product on [surface] in [environment]. Include [supporting objects] without distracting from the product.
Composition:
Use a [front / three-quarter / top-down] view. Position the product [placement] and leave clean negative space at [location].
Lighting:
Use [soft studio / natural window / dramatic directional] lighting with realistic shadows and controlled reflections.
Background:
[Describe color, texture, and depth.]
Brand mood:
[Minimal / luxurious / playful / sustainable / technical].
Requirements:
- Keep the product as the clear focal point.
- Preserve realistic proportions and materials.
- Do not add unsupported logos, labels, or text.
- Avoid distorted packaging and duplicated objects.
- Format: [orientation and aspect ratio].Infographic Brief
Generate a clean visual concept for a process, comparison, framework, or small set of verified data.
Create a clean infographic concept about [topic].
Communication goal:
[Describe what the viewer should understand.]
Audience:
[Describe the intended viewer.]
Information structure:
1. [Section or step 1]
2. [Section or step 2]
3. [Section or step 3]
4. [Section or step 4]
Layout:
Use a clear [vertical process / side-by-side comparison / circular framework / card grid] with strong visual hierarchy and generous spacing.
Visual system:
- Simple icons for each section
- Limited palette: [colors]
- High contrast
- Consistent shapes and line weights
- Minimal decorative elements
Requirements:
- Use only the supplied facts and values.
- Do not invent statistics.
- Keep text areas short and readable.
- Leave precise labels and final typography to be added in a design tool.
- Format: [orientation, dimensions, and aspect ratio].
Verified content:
[Paste labels, facts, and data]How to Improve a Prompt
Improve a weak prompt by replacing ambiguity with a clear outcome, relevant context, useful constraints, and a defined output format.
Write a blog post about cybersecurity.What is missing?
- No defined audience
- No specific topic or purpose
- No length or level of detail
- No tone or writing style
- No required structure
- No accuracy requirements
Act as a cybersecurity educator.
Write a beginner-friendly article explaining how small businesses can reduce phishing risk.
Audience:
Small-business owners with limited technical knowledge.
Requirements:
- Use plain English.
- Keep the article between 800 and 1,000 words.
- Explain what phishing is and how common attacks work.
- Include five practical prevention steps.
- Add a short employee-training checklist.
- Avoid unsupported statistics and exaggerated claims.
- Define unfamiliar technical terms.
Output format:
Use a clear title, short introduction, descriptive H2 headings, bullet points, and a concise conclusion with next actions.Why is it stronger?
- Defines the role and subject
- Identifies the target reader
- Sets scope and length
- Adds useful content requirements
- Controls accuracy and readability
- Specifies the final structure
Improve Any Prompt
Use this sequence when an AI response is too generic, inconsistent, or difficult to use.
-
01
Define the outcome
State exactly what the AI should create, analyze, or decide.
-
02
Add relevant context
Explain the audience, goal, situation, and available information.
-
03
Set useful constraints
Define scope, tone, length, requirements, and exclusions.
-
04
Specify the output
Describe the structure, fields, formatting, and success criteria.
-
05
Test and refine
Evaluate the response and revise only the instructions that need improvement.
Common Prompting Mistakes
Most poor AI responses come from missing direction, insufficient context, contradictory requirements, or an unnecessarily complicated request.
Vague Instructions
The task does not define what a successful response should contain.
Make this better.
Rewrite this introduction for clarity and concision. Keep the original meaning, use plain English, and limit it to 100 words.
Fix: Name the action, quality criteria, boundaries, and desired output.
Missing Context
The AI lacks information about the audience, goal, or situation.
Write an email about the delay.
Write a brief, apologetic email to a customer explaining that their order will arrive three days late. Offer a revised delivery date and a clear contact option.
Fix: Add only the background information needed to make the response accurate and relevant.
Conflicting Requirements
Two or more instructions cannot be satisfied at the same time.
Explain this comprehensively in one short sentence.
Begin with a one-sentence summary, then provide a detailed explanation under three descriptive headings.
Fix: Prioritize requirements or separate them into different parts of the response.
Too Many Tasks at Once
A single prompt combines unrelated deliverables and competing goals.
Research the market, create a strategy, write the website, design an ad campaign, and build a launch calendar.
Start by creating a market-research brief. Include the target audience, key competitors, customer needs, and unanswered questions. We will use the approved brief for the next stage.
Fix: Break complex work into stages with one clear deliverable at a time.
Structure, Accuracy, and Iteration
Good prompts provide enough control to make the result useful without adding irrelevant rules or assuming the first response will be perfect.
No Output Format
The requested information may be correct but difficult to use.
Compare these three project-management tools.
Compare the three tools in a Markdown table with columns for Price, Key Features, Limitations, Integrations, and Best Use Case.
Fix: Specify the structure, fields, length, and formatting needed for the next action.
Unnecessary Over-Prompting
Excessive rules distract from a simple task and may conflict.
Act as ten different experts, use twelve stages, explain every possible viewpoint, and provide twenty output formats for this two-sentence summary.
Summarize the text in two sentences for a general audience. Preserve the main conclusion and one important qualification.
Fix: Use the shortest instruction that communicates the goal and necessary constraints.
Missing Accuracy Rules
The prompt encourages a factual answer without controlling evidence.
Add convincing statistics and expert quotes to support the article.
Use only the supplied sources. Cite every statistic and quotation. If supporting evidence is unavailable, mark the claim as [SOURCE NEEDED].
Fix: Define permitted sources, citation rules, uncertainty handling, and what the AI should do when evidence is missing.
Expecting a Perfect First Response
The initial output is accepted without review or focused feedback.
Try again. This is not good.
Revise the draft by shortening the introduction, removing repeated ideas, adding one practical example, and preserving all verified facts.
Fix: Review the output against explicit criteria and give specific revision instructions.
Prompt Testing and Evaluation
Evaluate an AI response against explicit criteria instead of relying on whether it merely sounds convincing. Check both content quality and compliance with the original prompt.
Accuracy
Are factual claims correct, supported, and free from invented details?
Verify important claims independently.Relevance
Does the response directly address the requested task, audience, and goal?
Remove unrelated or distracting content.Completeness
Are all required questions, sections, fields, and deliverables included?
Compare the response with every requirement.Clarity
Is the response understandable, logically organized, and appropriate for the audience?
Check wording, structure, and unexplained terms.Format Compliance
Does the answer follow the requested length, structure, fields, and formatting?
Validate tables, JSON, headings, and limits.Consistency
Are terminology, tone, assumptions, values, and conclusions internally consistent?
Look for contradictions across the response.Source Grounding
Does the response remain within the supplied evidence and cite it correctly?
Trace claims back to permitted sources.Uncertainty Handling
Does the response identify assumptions, missing information, and limits of confidence?
Prefer transparent uncertainty over guessing.AI Response Evaluation Prompt
Score a response against the original prompt, identify specific weaknesses, and create a targeted revision.
Score and Improve an AI Response
Evaluate the AI response against the original prompt.
Use this scoring scale:
0 = Missing or incorrect
1 = Partially meets the criterion
2 = Fully meets the criterion
Score these criteria:
1. Accuracy
2. Relevance
3. Completeness
4. Clarity
5. Format compliance
6. Internal consistency
7. Source grounding
8. Uncertainty handling
Return:
1. A Markdown score table with Criterion, Score, Evidence, and Required Improvement
2. Total score out of 16
3. The three most important weaknesses
4. Any unsupported or unverifiable claims
5. Any missing requirements from the original prompt
6. A concise revision plan
7. A revised response that addresses the identified problems
Do not assume a claim is correct merely because it sounds plausible. If accuracy cannot be verified from the supplied information, mark it as "Needs verification."
Original prompt:
[Paste original prompt]
AI response:
[Paste AI response]
Permitted sources or reference material:
[Paste sources, or write "None supplied"]Prompt Engineering Best Practices
Use this checklist before, during, and after prompting to improve response quality without making every instruction unnecessarily long.
Before You Prompt
- Define the outcome. Know what the final response should help you understand, decide, create, or complete.
- Collect reliable input. Prepare the facts, source material, examples, and constraints the AI actually needs.
- Choose one primary task. Separate unrelated deliverables into different stages.
- Identify the audience. Decide who will read or use the result and what they already know.
While Writing the Prompt
- Lead with a clear instruction. Use a specific action verb and name the expected deliverable.
- Add only relevant context. Include background that materially changes the answer.
- Set useful constraints. Define scope, tone, length, requirements, exclusions, and sources.
- Specify the output format. State the required structure, fields, and formatting.
After You Receive the Response
- Check every requirement. Confirm that no requested section, field, or constraint was missed.
- Verify important claims. Check facts, calculations, quotations, sources, and technical details.
- Look for uncertainty. Identify assumptions, missing information, and unsupported confidence.
- Give focused feedback. Request specific revisions instead of asking the AI to “try again.”
Frequently Asked Questions
Quick answers to common questions about writing, testing, and improving prompts for generative AI tools.
What is prompt engineering?
Prompt engineering is the process of designing, testing, and refining instructions for an AI system. The goal is to provide enough task definition, context, constraints, and output guidance to produce a useful and reliable response.
What makes a good AI prompt?
A good prompt clearly states the task, supplies relevant context, defines important constraints, identifies the audience, and specifies the desired output format. Examples are useful when the required pattern, tone, or classification could be misunderstood.
How long should an AI prompt be?
A prompt should be only as long as necessary to communicate the task and requirements. Simple tasks may need one sentence, while complex work may require context, source material, constraints, examples, and a detailed output structure. Longer does not automatically mean better.
What is the difference between zero-shot and few-shot prompting?
Zero-shot prompting gives the AI an instruction without a demonstration. Few-shot prompting adds a small number of examples that show the desired input-and-output pattern. Few-shot prompting is especially useful for classification, consistent formatting, and style imitation.
Does assigning the AI a role improve accuracy?
A role can help establish a useful perspective, vocabulary, or level of detail, but it does not guarantee expertise or factual accuracy. Important claims should still be checked against reliable sources.
Can the same prompt be used with different AI tools?
The core structure of a good prompt is usually transferable across tools such as ChatGPT, Claude, Gemini, and Copilot. However, available features, context limits, tool access, formatting behavior, and supported parameters may differ, so some adaptation is often needed.
How can prompts reduce AI hallucinations?
Provide reliable source material, restrict the response to permitted sources, require citations for factual claims, ask the AI to identify missing information, and instruct it not to guess. These techniques reduce unsupported claims but cannot guarantee a completely accurate response.
Can a well-written prompt guarantee a correct answer?
No. A strong prompt can improve relevance, structure, and consistency, but it cannot guarantee correctness. Verify important facts, calculations, citations, code, and high-impact decisions independently.
Should I ask an AI to explain its reasoning?
Ask for a concise rationale, supporting evidence, assumptions, or verifiable calculation steps when they help you evaluate the answer. Focus on checkable justification rather than requesting an extensive hidden reasoning transcript.
Official Prompting Resources
Prompting behavior and supported features change over time. Use these official guides for current, tool-specific recommendations.