Masterclass · 2026

Prompt KING

Become a Prompt King 👑

The Complete Guide to Prompt Engineering — Every Framework, Technique, and Example A to Z

BasicsFrameworksAdvanced100+ Examples

Goal: After this, you'll master every major prompt framework, technique, and use case — and become a Prompt King across any AI tool.

Prompt King A to ZSlide 01 / 70

What You'll Master

🌱

Foundations

What is a prompt, how models work

📝

Core Principles

Clarity, specificity, iteration

🧩

Frameworks A-Z

CLEAR, RACE, RISEN, CRISPE + more

🎯

Techniques

CoT, few-shot, role, meta

🤖

Agentic Prompting

ReAct, plan-execute, tools

💼

Use Cases

Writing, coding, business, research

⚡

Power Moves

Chaining, self-critique, advanced

👑

King's Path

Library, workflow, mastery

Prompt King A to ZSlide 02 / 70

What is a Prompt?

A prompt is the instruction, question, or input you give to an AI model to elicit a desired response.

📝

Instruction

"Write a blog post about AI"

❓

Question

"What's the difference between Python 2 and 3?"

📦

Input Data

"Summarize this text: [paste]"

🎭

Role & Context

"Act as a teacher explaining to a 10-year-old"

📋

Output Spec

"Give me 5 bullets, under 50 words"

🎯

Few-Shot

"Example 1: X → Y. Example 2: X → Y. Now: X → ?"

Why it matters

AI doesn't read your mind. The quality of your input determines the quality of output. A great prompt is a clear contract between your intent and the AI's capabilities.

Prompt King A to ZSlide 03 / 70

How LLMs Think

LLMs predict the most likely next token. They don't truly "understand" — they recognize patterns from vast training data.

1

Tokenize

Your text is split into tokens

2

Encode

Tokens mapped to vectors

3

Context

Patterns matched against learned data

4

Predict

Next most likely token calculated

5

Generate

One token at a time

6

Sample

Temperature affects randomness

Implication

LLMs are next-token predictors. They don't "know" things — they generate plausible continuations. Specific prompts channel that prediction toward what you actually need.

Prompt King A to ZSlide 04 / 70

Anatomy of a Perfect Prompt

The 7 Components

Role: Who is the AI?
Task: What should it do?
Context: Background information
Format: How to structure output
Length: Word/character limits
Tone: Formal, casual, friendly
Examples: Sample inputs/outputs

Role: You are an expert Python developer Task: Write a function that validates email addresses Context: For a web app signup form Format: Python function with type hints Length: Under 30 lines Tone: Clean, production-ready code Examples: Reject emails with spaces, accept subdomains Now write the function.

Prompt King A to ZSlide 05 / 70

The 10 Golden Principles

Be Specific: Vague in, vague out
Show Examples: Don't tell, demonstrate
Set Context: Tell AI who it's writing for
Define Format: Tables, bullets, JSON
Constrain Length: "Keep under 100 words"

Assign Role: "You are an expert..."
Iterate: Refine based on outputs
Chain: Break complex tasks
Verify: AI can hallucinate
Test: Run prompts on different inputs

Prompt King A to ZSlide 06 / 70

Weak vs Strong Prompts

WEAK ❌ "Write about dogs."

STRONG ✅ "Write a 500-word blog post about training golden retrievers for first-time owners. Tone: friendly, practical. Include 3 common mistakes. End with a CTA to a training course."

WEAK ❌ "Make code for login"

STRONG ✅ "Write a Python Flask login endpoint with JWT authentication. Include: input validation, password hashing with bcrypt, error handling, and unit tests. Comment each section."

WEAK ❌ "Help me with marketing"

STRONG ✅ "Create 10 LinkedIn post hooks for our new AI analytics tool launch. Target: data engineers at mid-size companies. Tone: confident, not salesy. Under 80 chars each."

Prompt King A to ZSlide 07 / 70

Temperature, Tokens & Settings

🌡️

Temperature

0 = deterministic, 1 = creative

🪙

Tokens

~1 token = 4 chars or 0.75 words

📏

Max Tokens

Limits response length

🎯

Top-p

Nucleus sampling for diversity

⚡

Frequency

Penalty for repeated words

🔁

Presence

Penalty for repeating topics

Tip

Lower temperature (0.2) for factual/code tasks. Higher (0.8) for creative writing. Most prompts work fine with default 0.7.

Prompt King A to ZSlide 08 / 70

Prompt Frameworks — Your Recipes

Frameworks give you a repeatable structure. They remove the guesswork and ensure consistent quality.

📋

CLEAR

Concise, Logical, Explicit, Adaptive, Reflective

🎯

RACE

Role, Action, Context, Expectation

🌟

RISEN

Role, Instructions, Steps, End goal, Narrowing

🎨

CRISPE

Capacity/Role, Insight, Statement, Personality, Experiment

🔗

RTGO

Role, Task, Goal, Output

🏗️

CO-STAR

Context, Objective, Style, Tone, Audience, Response

Prompt King A to ZSlide 09 / 70

Framework #1: CLEAR

C - Concise: Keep it brief and direct L - Logical: Order your request sensibly E - Explicit: State exactly what you need A - Adaptive: Adjust based on first response R - Reflective: Iterate and improve Example: "Create a 10-step onboarding checklist for new SaaS users. Concise steps, one per line. Focus on activation metrics. Return as a numbered list."

When to use

Everyday tasks, content creation, simple questions. CLEAR is the universal starter framework.

Prompt King A to ZSlide 10 / 70

Framework #2: RACE

R - Role: "You are a senior content strategist" A - Action: "Write a 30-day content calendar" C - Context: "For a B2B SaaS selling to HR managers" E - Expectation: "Output as a table with columns: Date, Channel, Topic, CTA" Full example: "You are a senior content strategist at a Fortune 500 SaaS company. Write a 30-day LinkedIn content calendar. Context: We're launching an AI-powered recruitment tool targeting HR directors. Output as a markdown table with columns: Day, Topic, Format, CTA. Tone: thought leadership, not salesy."

When to use

Complex professional tasks — content, strategy, analysis. RACE gives structure without over-engineering.

Prompt King A to ZSlide 11 / 70

Framework #3: RISEN

R - Role: "You are an experienced product manager" I - Instructions: "Create user stories for..." S - Steps: "Follow these steps: 1) Identify user, 2) Define need, 3) Write story" E - End goal: "Goal is a sprint-ready backlog" N - Narrowing: "Keep under 200 words per story" Example: "You are an experienced product manager at a SaaS startup. Create 5 user stories for a new analytics dashboard feature. Follow this format: 'As a [user], I want [action], so that [benefit].' Include acceptance criteria. Goal is dev-ready stories. Keep each story under 100 words."

Prompt King A to ZSlide 12 / 70

Framework #4: CRISPE

C - Capacity/Role: Define the role R - Insight: Background context I - Statement: Your actual task P - Personality: Tone & style E - Experiment: Try variations Example: "Role: You are a senior B2B copywriter with 15 years at HubSpot. Insight: We've seen conversion rates drop 20% on landing pages. Task: Rewrite our homepage hero section. Personality: Friendly, confident, results-focused. Experiment: Provide 3 variations (bold, subtle, data-driven)."

When to use

Creative writing, marketing, and any task requiring a specific voice or multiple options to choose from.

Prompt King A to ZSlide 13 / 70

Framework #5: RTGO (Simple)

R - Role T - Task G - Goal O - Output format Example: "Role: Senior data scientist Task: Analyze this sales CSV for patterns Goal: Find top 3 growth opportunities Output: Bulleted list with metrics Perfect for quick tasks that don't need detailed setup."

When to use

Quick tasks, quick prompts, when you're in a hurry. The minimum viable prompt framework.

Prompt King A to ZSlide 14 / 70

Framework #6: CO-STAR (Singapore)

C - Context: Background O - Objective: What you want S - Style: Writing style T - Tone: Emotional tone A - Audience: Who reads it R - Response: Format Example: "Context: We're a fintech startup launching a new app. Objective: Write launch announcement blog post. Style: TechCrunch-style journalism. Tone: Excited but credible. Audience: Tech-savvy early adopters, age 25-40. Response: 800-word article with H2 subheadings, intro hook, and 3 key benefits bullets."

Winner of Singapore prompt engineering competition

CO-STAR is beloved by marketers and content creators for its balance of detail and clarity.

Prompt King A to ZSlide 15 / 70

Framework #7: TAG (Simplest)

T - Task A - Action G - Goal Example: "Task: Refactor this Python function. Action: Improve readability and add docstrings. Goal: Pass code review with senior team. The minimal framework. Perfect when you know what you want but need to clarify it."

Prompt King A to ZSlide 16 / 70

Framework #8: APE

A - Action P - Purpose E - Expectation Example: "Action: Write 5 cold emails Purpose: Sell SaaS to CTOs at Series A startups Expectation: Under 100 words, casual tone, clear CTA Great for marketing and sales tasks."

Prompt King A to ZSlide 17 / 70

Framework #9: Chain-of-Density (Advanced)

A special framework for generating increasingly dense, information-rich summaries.

Step 1: Summarize the article in 1-2 sentences. Step 2: Add 1-2 entities (names, dates, numbers) and condense. Step 3: Add 1-2 more entities and condense further. Step 4: Add 1-2 more entities and condense. Step 5: Final summary should be dense, info-rich, no more than 80 words. Article: [paste article]

Best for

News articles, research papers, long documents — when you want maximum information density.

Prompt King A to ZSlide 18 / 70

Framework #10: Chain-of-Thought

Force the AI to think step-by-step. Massively improves reasoning, math, and logic.

"Solve this step by step. Show each step clearly. A store sells 120 shirts daily at $25 each. Fixed costs are $1,500/day. Variable cost is $12 per shirt. What's the daily profit? Step 1: Calculate revenue Step 2: Calculate variable costs Step 3: Calculate total costs Step 4: Calculate profit Show your work for each step."

Magic phrase

Add "Let's think step by step" or "Show your reasoning" — dramatically improves complex answers.

Prompt King A to ZSlide 19 / 70

Technique #1: Few-Shot Prompting

Show examples of input-output pairs. AI learns the pattern and applies it.

"Translate English to French: English: Hello → French: Bonjour English: Goodbye → French: Au revoir English: Thank you → French: Merci English: Good morning → French:"

Why it works

Examples teach the model the exact pattern, format, and style you want. Use 2-5 examples for best results.

Prompt King A to ZSlide 20 / 70

Technique #2: Zero-Shot Prompting

No examples — just direct instruction. Works for simple, well-known tasks.

"Translate 'Good morning' to French."

"Classify this review as Positive, Neutral, or Negative: 'Amazing product, would buy again!'"

"Write a haiku about autumn."

"What is the capital of France?"

⚡ Best for simple, common tasks. Use Few-Shot for nuanced or unusual requirements.

Prompt King A to ZSlide 21 / 70

Technique #3: Role Prompting

🎓

Expert

"You are a Harvard professor..."

🎨

Creative

"You are a poet from the 1800s..."

💻

Technical

"You are a senior Python dev..."

🔍

Analyst

"You are a data scientist skeptical of correlation..."

💼

Business

"You are a McKinsey consultant..."

🧘

Coach

"You are a patient executive coach..."

Pro tip

Be specific. "Marketing expert" beats "smart person". Add years of experience or specific achievements.

Prompt King A to ZSlide 22 / 70

Technique #4: System Prompts

Set persistent instructions that apply to every message in the conversation.

System: You are a senior financial analyst. Always: - Cite your sources - Show your work - Use tables for comparisons - Flag assumptions - End with a "Risks" section User: Should I invest in Tesla stock?

Where to use

ChatGPT Custom Instructions, Claude Projects, API system parameter, Cursor AI rules.

Prompt King A to ZSlide 23 / 70

Technique #5: Negative Prompting

Tell AI what NOT to do. Equally important as telling it what to do.

"Write a professional email. Do NOT: - Use jargon or buzzwords - Include emojis - Start with 'I hope this email finds you well' - Exceed 150 words - Use exclamation marks Just be direct and helpful."

Best for

Image generation (negative prompts are huge), style control, preventing common AI mistakes.

Prompt King A to ZSlide 24 / 70

Technique #6: Self-Critique

"Write a 200-word product description for a wireless headphone. Then critique your own response: 1. Is it engaging? 2. Are benefits clear? 3. Any clichés? 4. Better alternatives? Then provide an improved version."

Use case

When you want self-improving outputs. AI often catches mistakes it would otherwise make.

Prompt King A to ZSlide 25 / 70

Technique #7: Meta Prompting

Use AI to write better prompts. The ultimate force multiplier.

"I want to ask an AI to help me plan a team offsite. Improve my prompt to get a more detailed and actionable response: My current prompt: 'Help me plan a team offsite.' Provide 3 improved versions with different approaches, explaining the techniques you used."

Why it works

AI knows what works well in prompts. Use it to optimize your prompts before sending.

Prompt King A to ZSlide 26 / 70

Technique #8: Reverse Prompting

"Here is an output I admire: [example output] What prompt could have generated this? Give me the exact prompt to reuse."

Use case

When you see good AI outputs and want to recreate the magic. Perfect for studying examples and building prompt libraries.

Prompt King A to ZSlide 27 / 70

Technique #9: Devil's Advocate

"My plan: Launch a $50/month AI tool for solo lawyers. First, argue strongly FOR this plan. Then, argue strongly AGAINST this plan. Then, give me your balanced recommendation. Step 1: Argue FOR [topic] Step 2: Argue AGAINST [topic] Step 3: Balanced view with key trade-offs"

Use case

Strategic decisions, plan validation, considering opposing views. Excellent for executives.

Prompt King A to ZSlide 28 / 70

Technique #10: Iterative Refinement

First draft is rarely the best. Refine through structured iterations.

Iteration Chain

1st: "Draft a LinkedIn post about AI"
2nd: "Make it more engaging"
3rd: "Add a personal story hook"
4th: "Shorten to 150 words"
5th: "Add 3 relevant hashtags"
6th: "Make the CTA stronger"

Refinement Phrases

"Make it more..."
"Add an example of..."
"Simplify this for..."
"Rewrite in the style of..."
"Make the opening stronger"
"Expand section X"

Prompt King A to ZSlide 29 / 70

Technique #11: Prompt Chaining

Break complex tasks into a sequence of smaller prompts. Each output feeds the next.

Step 1: "Generate 10 blog topics about [industry]" Step 2: "Pick top 3 and create outlines for each" Step 3: "Write a 500-word draft for topic #1" Step 4: "Generate 5 social media posts from this blog" Step 5: "Suggest 10 SEO keywords for this blog"

Why chain

Each step is focused, easier to debug, and the final output is much higher quality.

Prompt King A to ZSlide 30 / 70

Agentic #1: ReAct (Reason + Act)

1

Reason

Think about what to do

2

Act

Use a tool or take action

3

Observe

See the result

4

Repeat

Until goal reached

"You can use tools. For each step: Thought: What you need to do Action: tool_name({args}) Observation: result Goal: Find the top 5 customers by revenue in 2024. Start."

Prompt King A to ZSlide 31 / 70

Agentic #2: Plan-and-Execute

"Goal: Plan a 5-day trip to Japan for $2000. Step 1: Create a detailed plan with daily activities, budget breakdown, and bookings needed. Step 2: Execute the plan step by step. PLAN: 1. Flights (book) 2. Hotel (book) 3. Day 1: Tokyo 4. Day 2: Kyoto 5. Day 3: Osaka 6. Day 4: Nara 7. Day 5: Tokyo 8. Budget tracker Execute now."

Prompt King A to ZSlide 32 / 70

Agentic #3: Reflection

"Solve this customer complaint. Then reflect on your response. Customer: 'Your software keeps crashing!' Your response: [draft response] Reflection: - Did I acknowledge the issue? - Did I offer a solution? - Did I sound empathetic? - Score: X/10 - Improved version:"

Use case

Quality-critical outputs. AI improves itself iteratively when asked to reflect.

Prompt King A to ZSlide 33 / 70

Agentic #4: Multi-Agent Prompting

"You have 4 agents: Agent 1 (Strategist): Creates the overall plan Agent 2 (Researcher): Provides data and facts Agent 3 (Critic): Identifies weaknesses Agent 4 (Writer): Drafts final output Collaborate to create a SaaS launch plan. Round 1: Strategist proposes Round 2: Researcher adds data Round 3: Critic critiques Round 4: Writer drafts Continue until complete."

Prompt King A to ZSlide 34 / 70

Advanced: XML Tag Prompting

Use XML tags to structure complex prompts. Claude especially responds well to this.

<role>Senior product manager</role> <context>B2B SaaS for HR teams</context> <task>Write product requirements document</task> <constraints> - Under 1000 words - Use bullets for features - Include success metrics </constraints> <output> # Overview # User Stories # Features # Metrics </output>

Prompt King A to ZSlide 35 / 70

Advanced: Markdown Structure

"# Task Create a quarterly business review. # Context We are a B2B SaaS with 500 customers. # Requirements - Executive summary (5 bullets) - Revenue breakdown by segment - Top 3 wins and 3 challenges - Q4 priorities # Format Use tables for numbers. Keep under 1500 words."

Prompt King A to ZSlide 36 / 70

Advanced: Token-Efficient Prompts

Use abbreviations for repeated words: "img" for image
Replace full sentences with structured bullets
Cut filler: "I want you to..." → just give the task
Reference previous outputs: "Same as before but for X"
Use numbers: "3 reasons" beats "give me three reasons"
Drop politeness: AI doesn't need "please" or "thank you"

Prompt King A to ZSlide 37 / 70

Use Cases: Writing

"Write a 1500-word blog post about 'The Future of Remote Work' for HR managers at mid-size companies. Use: hook story, 3 main benefits with stats, 2 challenges with solutions, hopeful conclusion. Tone: authoritative but warm."

"Rewrite this email to be: - Under 100 words - Friendly tone - Clear ask - Pinned action item at end Email: [paste]"

"Generate 10 LinkedIn post hooks about [topic]. Target audience: [audience]. Each hook under 80 chars. Use curiosity, not clickbait."

"Create an SEO-optimized blog outline for '[keyword]'. Include: H2 subheadings, FAQ section, meta description under 160 chars, target word count 1500-2000."

Prompt King A to ZSlide 38 / 70

Use Cases: Coding

"Write a Python function that validates email addresses. Include: regex pattern, edge case handling, type hints, docstring, 5 unit tests using pytest."

"Debug this JavaScript error: [error message]. Here's my code: [paste code]. Explain the cause and provide a fixed version with comments."

"Refactor this React component for better performance: - Use React.memo where appropriate - Avoid unnecessary re-renders - Split into smaller components if needed [paste component]"

"Convert this SQL query from MySQL to PostgreSQL syntax, handling: data types, functions, pagination, JSON columns. Query: [paste query]"

Prompt King A to ZSlide 39 / 70

Use Cases: Business & Strategy

"Create a SWOT analysis for entering the AI education market in India. Include: realistic strengths/weaknesses, market data, competitor landscape, actionable recommendations."

"Write a 5-page investor pitch deck outline for our Series A. Include: problem, solution, market size, traction, business model, team, ask. Format as slide titles + 3 bullets each."

"Design a customer feedback survey for our SaaS product. 10 questions max. Mix: rating scales, multiple choice, and 1 open-ended. Goal: measure NPS and feature satisfaction."

"Generate 5 PR strategy angles for our product launch. Target: tech journalists. Each angle: hook, supporting data, target publication."

Prompt King A to ZSlide 40 / 70

Use Cases: Research & Analysis

"Research [topic] and provide 10 key findings with credible source links. Last updated: 2026. Flag any conflicting opinions."

"Analyze this CSV and tell me: - Top 3 trends - 2 anomalies - 1 actionable insight Data: [paste]"

"Summarize this 50-page report in: - 5 key findings - 3 recommendations - 1 main risk - Under 500 words Report: [paste]"

"Compare these two options across: cost, features, ease of use, support. Format as a table with a winner for each category. Option A: [paste] Option B: [paste]"

Prompt King A to ZSlide 41 / 70

Use Cases: Marketing & Sales

"Write a 5-email welcome sequence for new SaaS users. Goal: activation by day 7. Tone: helpful, not salesy. Each email: subject, preview, body under 200 words."

"Create 20 LinkedIn hooks for our new AI feature. Target: data scientists. Tone: technical, not marketing-speak. Each under 80 chars."

"Generate 30 SEO keywords for [topic]. Include: short-tail, long-tail, question-based. Group by search intent (informational, commercial, transactional)."

"Write a sales email to a CTO at a Series B startup. Goal: book a 15-min demo. Subject + body under 150 words. Casual, value-focused, no corporate jargon."

Prompt King A to ZSlide 42 / 70

Use Cases: Education & Learning

"Explain quantum computing like I'm a 12-year-old. Use an analogy from everyday life. Keep under 300 words. End with one curiosity question."

"Create a 30-day study plan for learning Python. Daily: 1 hour. Include: topics, resources, practice projects, milestones. Format as a table."

"Generate 20 flashcards on [topic]. Format: front (question) / back (concise answer). Focus on high-yield exam topics."

"I'm teaching [topic] to beginners. Create a 60-minute lesson plan with: hook (5 min), concept (15 min), example (15 min), practice (20 min), wrap-up (5 min)."

Prompt King A to ZSlide 43 / 70

Use Cases: Personal & Lifestyle

"Plan a 7-day trip to [destination] with $2000 budget for 2 people. Include: flights, hotels, daily itinerary, restaurant recommendations, hidden gems. Format: day-by-day table."

"Create a 4-week beginner workout plan with no equipment. 30 min daily. Mix: strength, cardio, flexibility. Include rest days. Provide video links."

"Generate a weekly meal plan for [diet preference] with a shopping list. Budget: $[amount]. Serves [number]. Focus on variety and nutrition."

"Help me write a professional resignation letter. Reason: [your reason]. Tone: gracious, professional, leaves bridges unburned. 2 weeks notice."

Prompt King A to ZSlide 44 / 70

Advanced Power Moves

🪞

Perspective Shift

"Answer as if you were Steve Jobs"

🧪

A/B Test

"Give me 3 different approaches"

🔍

Critique Mode

"Find 5 weaknesses in this plan"

🌐

Translate Tone

"Make this sound like Yoda"

⏳

Time Travel

"Predict 10 years from now"

🎯

Steelman

"Argue the strongest case for the opposite view"

Prompt King A to ZSlide 45 / 70

Advanced: Persona Stacking

"You are simultaneously: - A senior software architect - A cybersecurity expert - A DevOps engineer - A product manager Review this system design from all four perspectives. Identify: 1. Technical concerns (architect) 2. Security vulnerabilities (cybersecurity) 3. Deployment risks (DevOps) 4. User experience issues (PM) System: [paste design]"

Best for

Complex decisions needing multiple perspectives. Better than asking for one opinion.

Prompt King A to ZSlide 46 / 70

Advanced: Output Format Control

Tables

"Compare X, Y, Z as a table with columns: Feature, Pros, Cons, Price"

JSON

"Return valid JSON: {product, price, features}"

Markdown

"Format with H2 headings, bold for key terms"

Email

"Format as: Subject, Greeting, Body, Sign-off"

Code

"Return only code in markdown code blocks, no prose"

Custom Schema

"Output as: ## Header, then 3 bullets, then a quote"

Prompt King A to ZSlide 47 / 70

Image Generation Prompts

"A photorealistic portrait of an elderly Japanese fisherman, weathered skin, intense gaze, golden hour lighting, shot on Canon EOS R5, 85mm lens, shallow depth of field, cinematic color grading Negative prompt: blurry, low quality, distorted, watermark, text"

Image prompt formula

Subject + Style + Lighting + Composition + Camera + Mood + Negative

Prompt King A to ZSlide 48 / 70

Tool / Function Calling Prompts

"Available tools: - search_web(query) - get_weather(location) - calculate(expression) - send_email(to, subject, body) User: 'What's the weather in Tokyo and email it to boss@company.com' Call appropriate tools in order. Stop when done."

Use case

AI agents, automation, API integration. Tell AI which tools exist and when to use them.

Prompt King A to ZSlide 49 / 70

Hallucination Defense Prompts

"Answer my question. Then add: - Confidence (High/Medium/Low) - Sources used (cite URLs) - Assumptions made - What you're uncertain about Question: [your question]"

Anti-hallucination

Add: "If unsure, say so. Don't make up information." Forces honesty over confidence.

Prompt King A to ZSlide 50 / 70

Common Mistakes

❌ Mistakes

Too vague prompts
No examples for nuanced tasks
One-shot only — no iteration
Copy-paste without review
Trusting AI facts blindly
Ignoring negative prompts
Using the same prompt everywhere

✅ Fixes

Add role + context + format + length
Include 2-5 examples
Iterate with refinement prompts
Always review and edit
Verify facts with sources
Use negative prompts to prevent errors
Maintain a personal prompt library

Prompt King A to ZSlide 51 / 70

Building Your Prompt Library

What to Store

Prompt name and purpose
Full template with placeholders
Example input and output
Best model/settings
Known limitations
Last tested date and version

How to Organize

By function: writing, coding, business
By role: marketing, sales, engineering
By frequency: daily, weekly, project
By model: GPT, Claude, Gemini

Prompt King A to ZSlide 52 / 70

Template Variables

Use placeholders to make reusable templates.

"Write a LinkedIn post about {{TOPIC}} for {{AUDIENCE}}. Key points: 1. {{POINT_1}} 2. {{POINT_2}} 3. {{POINT_3}} Tone: {{TONE}} Length: {{LENGTH}} words CTA: {{CTA}}"

Tools for prompt management

Notion, Airtable, Google Sheets, dedicated tools like PromptLayer, Helicone.

Prompt King A to ZSlide 53 / 70

Model-Specific Tips

💬

ChatGPT

Loves "Let's think step by step". Great with Custom Instructions.

🧠

Claude

Excels at XML tags & structured prompts. Natural for long docs.

🔷

Gemini

Best with multimodal inputs. Great for real-time info.

🔍

Perplexity

Built for search. Ask direct questions, get cited answers.

🌐

Open Source

Be more explicit. Provide more examples. Test temperature.

🤖

Agents

Use ReAct pattern. Define tools clearly. Plan first.

Prompt King A to ZSlide 54 / 70

Image AI Prompt Templates

Midjourney / DALL-E

"[Subject], [style], [lighting], [composition], [camera], --ar 16:9 --v 7"

Logo Design

"Minimalist logo for [brand], [colors], flat design, vector, white background"

Photorealistic

"Photorealistic [scene], shot on Sony A7III, 50mm lens, 4K, natural lighting"

Illustration

"Flat vector illustration of [concept], modern, vibrant colors, minimalist"

Prompt King A to ZSlide 55 / 70

Video AI Prompts

"Generate a 30-second product video showing [product]: Scene 1 (5s): Problem - person frustrated with [issue] Scene 2 (5s): Discovery - they find our product Scene 3 (10s): Use - showing key features Scene 4 (5s): Result - happy customer Scene 5 (5s): CTA - logo + tagline Style: cinematic, bright, modern Music: upbeat indie pop"

Prompt King A to ZSlide 56 / 70

Safety in Prompts

✅ Do

Add "Don't make up info" to fact prompts
Specify "Use age-appropriate language"
Include "Respect privacy" for personal data
Ask for sources in research

⚠️ Avoid

Sharing PII in prompts
Asking for harmful content
Trick prompts to bypass safety
Sharing confidential company data with public AI

Prompt King A to ZSlide 57 / 70

Prompt Versioning & Testing

Treat prompts like code. Version, test, and track what works.

🏷️

Version Naming

blog-generator-v1.0, blog-generator-v1.1

🧪

A/B Testing

Run two versions on same input, compare outputs

📊

Metrics

Track: output quality, time saved, errors

📝

Notes

What worked, what failed, why changed

📁

Repository

Git for prompts — track changes over time

🔄

Reviews

Have teammates review important prompts

Prompt King A to ZSlide 58 / 70

Evaluating Prompt Quality

Quality Metrics

Accuracy: Is it factually correct?
Relevance: Does it address the prompt?
Format: Does it follow format?
Tone: Is it the right voice?
Completeness: Is anything missing?
Conciseness: Is it tight or rambling?

Test Methods

Run same prompt 3-5 times — variance check
Test edge cases and unusual inputs
Have humans rate outputs blindly
Compare against golden examples
Track failure modes

Prompt King A to ZSlide 59 / 70

50+ Master Prompts to Copy

"Act as [ROLE]. I'm working on [TASK]. The audience is [WHO]. Format as [STRUCTURE]. Length: [WORD_COUNT]. Tone: [VOICE]. Start with a hook. End with a clear next step."

"Create a [DOCUMENT_TYPE] for [PURPOSE]. Include: 1) [SECTION_1], 2) [SECTION_2], 3) [SECTION_3]. Use [TONE] tone. Audience: [AUDIENCE]."

"I'm trying to [GOAL]. Current situation: [CONTEXT]. What's the 80/20 approach? List 3 quick wins and 1 long-term play."

"You are a [EXPERT_ROLE]. Review my [DELIVERABLE] and provide 5 specific, actionable improvements with examples."

"Explain [CONCEPT] to a [AUDIENCE_LEVEL]. Use 1 analogy and 1 real example. End with one practical application."

"Generate [NUMBER] [ITEMS] for [PURPOSE]. Each: [SPEC]. Ensure variety and originality. Rank by impact."

Prompt King A to ZSlide 60 / 70

50+ Master Prompts Continued

"Solve this problem step by step. State each step. Verify at each stage. Problem: [PROBLEM]"

"Create a checklist for [TASK]. Cover: preparation, execution, follow-up. Add time estimates. Format as numbered list."

"Translate this content for [AUDIENCE]. Keep the core message but adjust tone, examples, and complexity. Original: [TEXT]"

"I'm about to [ACTION]. What could go wrong? List 5 risks with mitigation strategies."

"Analyze this [DATA/DOCUMENT]. Identify: 3 patterns, 2 anomalies, 1 actionable insight. Format as brief report."

"Roleplay a [SCENARIO]. I'll play the [ROLE]. Stay in character. Provide feedback after each turn."

Prompt King A to ZSlide 61 / 70

Designing Prompt Workflows

1

Trigger

Event or input

2

Prompt 1

First processing

3

Prompt 2

Refine output

4

Format

Structure result

5

Deliver

Send or save

Example: Customer Email Workflow

Incoming email → AI classifies → AI drafts response → Human reviews → Sends out. 80% time saved.

Prompt King A to ZSlide 62 / 70

Tools for Prompt Management

📝

Notion

Prompt libraries & wikis

📊

Airtable

Structured prompt databases

💻

GitHub

Version control for prompts

🔧

PromptLayer

Prompt versioning & testing

📈

Helicone

LLM observability

🧪

LangSmith

LangChain debugging & testing

Prompt King A to ZSlide 63 / 70

Speed Mastery

Keyboard Shortcuts

Cmd/Ctrl + Shift + O — New chat
Cmd/Ctrl + K — Search history
Cmd/Ctrl + / — Show shortcuts
Shift + Enter — New line
Tab — Accept suggestion

Time-Savers

Pin best prompts to top of chat
Use voice input instead of typing
Save common prompts as snippets
Batch similar requests
Set up custom instructions once
Use Projects for recurring work

Prompt King A to ZSlide 64 / 70

Advanced Reasoning Techniques

Tree of Thoughts

"Generate 3 different solutions to this problem. Evaluate each. Pick the best. Refine."

Self-Refine

"Draft a response. Then critique it. Then provide an improved version. Repeat 2-3 times."

Constitutional AI

"Apply these principles: [list]. Check your response. Revise if violating principles."

Constitutional Chain

"Step 1: Reason. Step 2: Critique per rules. Step 3: Revise. Step 4: Final answer."

Prompt King A to ZSlide 65 / 70

Prompt Engineering as a Career

💼

Roles

Prompt Engineer, AI Engineer, AI PM, AI Content Lead

💰

Salary

$80K-$300K+ depending on seniority

📚

Skills to Build

LLM understanding, domain expertise, communication

🎯

Where to Work

Startups, agencies, enterprises, freelancing

📈

Career Path

Prompt Eng → AI Eng → AI Architect

🌟

Build Public Profile

Twitter, blog, GitHub, prompt libraries

Prompt King A to ZSlide 66 / 70

Common Questions

❓ Q: Should I learn prompting or coding?

A: Both. Prompting is the interface; coding unlocks automation. Start with prompting.

❓ Q: Will prompting be obsolete?

A: No — as models improve, prompts get MORE important, not less. It's the human-AI interface.

❓ Q: Which model is best?

A: Depends. GPT-5 for general, Claude for long docs, Gemini for Google, Perplexity for search.

❓ Q: How long to master prompting?

A: 2-4 weeks for basics, 3-6 months for advanced with daily practice.

Prompt King A to ZSlide 67 / 70

30-Day Prompt Mastery Plan

Week 1 · Foundations

Learn CLEAR, RACE, CO-STAR
Practice 10 prompts daily
Master role + format

Week 2 · Techniques

Few-shot, CoT, self-critique
Image and code prompts
Refinement workflows

Week 3 · Domains

Build prompts for your work
Marketing, coding, research
Create personal library

Week 4 · Advanced

Agentic patterns (ReAct)
Test in production
Share & teach others

Day 7

Comfortable with frameworks

Day 30

Building AI workflows confidently

Prompt King A to ZSlide 68 / 70

20 King's Rules

Be specific, not vague
Always provide context
Show examples for nuanced tasks
Specify output format
Use role + task pattern
Iterate, never settle for first draft
Verify facts independently
Chain complex prompts
Use XML/Markdown structure
Maintain a prompt library

Test with edge cases
Version your prompts
Use negative prompts
Match temperature to task
Self-critique for quality
Think step-by-step for logic
Use personas wisely
Protect sensitive data
Stay curious and experiment
Share knowledge with community

Prompt King A to ZSlide 69 / 70

Final · Slide 70 / 70

You Are Now the Prompt King

From foundations to advanced frameworks, from basic chat to agentic workflows — you now have every framework, technique, and example to master prompting.

🧠

Frameworks

CLEAR, RACE, CO-STAR, RISEN, CRISPE & more

🎯

Techniques

Few-shot, CoT, role, meta, self-critique

🤖

Agentic

ReAct, plan-execute, reflection, multi-agent

💼

All Use Cases

Writing, coding, business, research, marketing

⚡

Power Moves

Persona stacking, XML, chaining, evaluation

👑

King's Mindset

Build, share, iterate, master — every day

🎬 If this helped, like, share, subscribe!
💬 What's your favorite prompt framework? Comment below!

Prompt King A to Z · YouTube MasterclassSlide 70 / 70