Cheatsheet

Prompt Engineering Cheatsheet

Role Prompts

You are a senior [role] with 10 years of experience in [domain]...
Act as a [profession] who specializes in [area] and is known for [trait]...
Respond as if you are explaining this to a [audience level: beginner/expert/executive]...
You are a [role] at a [company type]. Your goal is to [objective]...
Adopt the perspective of a [role] who disagrees with the common approach to [topic]...
You are a world-class [role]. Think carefully before responding. Your reputation depends on accuracy.
You are a [role] preparing a [deliverable] for [audience]. Prioritize [key value: clarity/brevity/depth].

Output Formats

Respond in a markdown table with columns: [col1], [col2], [col3]
Give me a numbered list of exactly N items, ranked by [criteria]
Format your response as JSON with these keys: { title, summary, items[], score }
Write in bullet points, max 2 sentences each, no preamble
Structure your response with H2 headers for each section
Return a CSV with headers: [header1], [header2], [header3]
Write as a decision matrix comparing [options] across [criteria]
Format as a step-by-step checklist with checkboxes: - [ ] Step 1...

Constraint Patterns

Keep your response under [N] words
Use only simple language (8th grade reading level)
Do not include any technical jargon — explain terms if you must use them
Include at least 2 real-world examples from [industry]
Do not use filler phrases like 'In conclusion' or 'It is important to note'
Every claim must include a specific example or data point
Respond in [language]. Do not mix languages.
Avoid lists — write in flowing paragraphs only

Reasoning Patterns

Think step by step before answering
List the pros and cons before making a recommendation
Show your reasoning, then give the final answer on a separate line
Consider 3 different approaches, then recommend the best one with justification
First identify the key assumptions, then analyze each one
Play devil's advocate: argue the opposite position before giving your real answer
Break this problem into sub-problems, solve each, then synthesize
Rate your confidence (low/medium/high) and explain why

Advanced Techniques

System prompts: Set persistent instructions at the start — 'You always respond in British English and cite sources'
Temperature guidance: Use low temperature (0-0.3) for factual/analytical tasks, higher (0.7-1.0) for creative/brainstorming
Few-shot examples: Provide 2-3 input/output pairs before your actual request to 'teach' the desired format
Chain-of-thought: Add 'Let's think step by step' to improve reasoning on complex problems
Self-consistency: Ask the AI to solve the problem 3 times and pick the most common answer
Instruction hierarchy: Put the most important instruction first and last — models pay more attention to these positions
Meta-prompting: Ask the AI to write a better prompt for your task — 'What additional information would you need to give a better answer?'

Profession-Specific Prompts

Healthcare: 'You are a clinical researcher. Summarize this study for a patient audience. Flag any claims that lack strong evidence.'
Legal: 'You are a corporate attorney. Review this contract clause and identify risks. Do not provide legal advice — flag areas for human review.'
Marketing: 'You are a growth marketer. Write 5 variations of this ad headline targeting [audience]. Optimize for click-through rate.'
Engineering: 'You are a senior software engineer. Review this code for bugs, performance issues, and security vulnerabilities. Explain each finding.'
Finance: 'You are a financial analyst. Summarize this earnings report in 3 bullet points for an executive audience. Highlight risks.'
Education: 'You are a curriculum designer. Create a 4-week lesson plan for teaching [topic] to [grade level]. Include assessments.'

Debugging Bad Outputs

Output too generic? → Add more specific context about your use case, audience, and constraints
Output too long? → Add 'Be concise. Maximum [N] words' or 'Respond in 3 bullet points only'
Output hallucinating facts? → Add 'Only include information you are confident about. Say I don't know if unsure'
Output ignoring instructions? → Move the ignored instruction to the very end of the prompt (recency bias)
Output in wrong format? → Provide a concrete example of the exact format you want
Output too surface-level? → Add 'Go deeper. Explain the why, not just the what. Include specific examples'
Output inconsistent across runs? → Add few-shot examples and reduce temperature to 0
Output mixing up context? → Break into smaller, focused prompts instead of one mega-prompt

Prompt Templates

Content Brief: 'Write a content brief for a [type] about [topic]. Include: target audience, key points to cover, tone, word count, SEO keywords, and a suggested outline.'
Email Draft: 'Draft a [tone] email to [recipient type] about [topic]. The goal is to [desired outcome]. Keep it under [N] words. Include a clear CTA.'
Document Summary: 'Summarize this [document type] in [N] bullet points. Focus on: key decisions, action items, and open questions. Audience: [who will read this].'
Competitive Analysis: 'Compare [product A] vs [product B] for [use case]. Structure as a table with columns: Feature, Product A, Product B, Winner. Add a final recommendation.'
Meeting Prep: 'I have a meeting about [topic] with [attendees]. Generate: 5 talking points, 3 potential objections with responses, and 2 questions I should ask.'
Learning Plan: 'Create a [duration] learning plan for mastering [skill]. Break into weekly milestones with specific resources. Assume [current level] starting knowledge.'

Quick Reference

This cheatsheet covers the most essential prompt engineering patterns. Keep it handy when working with any AI chatbot — Claude, ChatGPT, Gemini, or any other LLM.

The CRISP Framework

The CRISP framework is a simple, repeatable structure for writing effective prompts. Use it as your default starting point for any AI interaction.

Context — Set the scene. What is the situation, background, or domain? Example: "We are a B2B SaaS company selling project management software to mid-market companies."
Role — Assign the AI a specific persona. This shapes the tone, vocabulary, and perspective of the output. Example: "You are a senior content strategist with 10 years of experience in SaaS marketing."
Instructions — State what you want clearly and directly. Use imperative verbs: "Write," "Analyze," "Compare," "List." Avoid vague requests like "help me with" or "tell me about."
Specifics — Add constraints that shape the output: word count, format, tone, audience level, number of items, what to include or exclude.
Proof — Ask the AI to show its work: provide reasoning, cite examples, or explain trade-offs. This reduces hallucination and increases output quality.

Prompt Anatomy

Every great prompt has these elements, even if some are implicit:

[Role] You are a ...
[Context] Background information the AI needs
[Task] What you want it to do
[Format] How you want the output structured
[Constraints] Length, tone, audience, exclusions
[Examples] Optional: show what good output looks like

Common Mistakes

Mistake	Fix
Too vague ("write something about marketing")	Add specific constraints: topic, audience, length, format
Too long and rambling	Break into multiple focused prompts
No examples provided	Add 1-2 few-shot examples of desired output
No format specified	Specify: bullet points, JSON, table, numbered list
Asking multiple unrelated questions at once	One task per prompt for best results
Not specifying audience	Add "for a [technical/non-technical/executive] audience"
Ignoring output quality issues	Ask the AI to self-critique: "Review your answer for accuracy"
Using AI for tasks that need real-time data	Pair with search tools or RAG for current information

Quick Decision Tree: Which Technique to Use?

Need a specific persona or expertise? → Role Prompt
Need structured data? → Output Format (JSON, table, CSV)
Need the AI to reason through a problem? → Chain-of-Thought
Need consistent outputs across similar tasks? → Few-Shot Examples
Need the AI to avoid certain topics or formats? → Constraint Patterns
Need complex multi-step analysis? → Step-by-Step Decomposition
Need the AI to consider multiple perspectives? → Multi-Perspective Prompt
Getting generic or shallow answers? → Specificity + Follow-up Prompts

Learn more: What is Prompt Engineering? | Write Your First AI Prompt

Get AI Tips Every Week

Get smarter about AI every week — practical tips, prompts, and workflows in your inbox.