Prompt Engineering Academy

Prompt Engineering Academy

Tech Horizon Academy teaches a small set of high-leverage prompting techniques. Pick the one that matches your task; mix them as you get more comfortable.

Zero-Shot Prompting (Beginner)

Give direct instructions without examples

• When to use: Simple tasks, clear expectations
• Why it works: Fast, efficient for straightforward requests
• Example: "Write a professional email declining a meeting invitation"
• Tip: Be specific about format and length expected

Few-Shot Prompting (Beginner)

Provide 2-3 examples to guide the output

• When to use: Pattern matching, consistent formatting, style replication
• Why it works: Shows the AI exactly what you want
• Example: "Here are 3 example product descriptions... Now write one for [product]"
• Tip: Choose diverse examples that cover edge cases

Chain-of-Thought (Intermediate)

Ask the AI to show its reasoning step-by-step

• When to use: Complex problems, analysis, decision-making
• Why it works: Improves accuracy by forcing logical progression
• Example: "Think through this step by step: Should we expand to Melbourne?"
• Tip: Add "Let's think step by step" or "Walk through your reasoning"

Constraint-Based (Intermediate)

Define boundaries, rules, and limitations

• When to use: Compliance requirements, specific formats, safety
• Why it works: Ensures output meets specific criteria
• Example: "Write a response that: uses no more than 100 words, includes an apology, offers a discount"
• Tip: List constraints clearly with bullet points

Persona Prompting (Intermediate)

Ask the AI to adopt a specific role or expertise

• When to use: Need expert perspective, professional advice
• Why it works: Frames responses from a knowledgeable viewpoint
• Example: "You are a seasoned Australian tax accountant. Explain..."
• Tip: Include the persona's experience level and expertise area

First Principles (Advanced)

Break problems down to fundamental truths

• When to use: Strategic decisions, innovation, challenging assumptions
• Why it works: Gets past surface-level thinking
• Example: "From first principles, what are the core needs our customers have?"
• Tip: Ask "What is the fundamental truth here?" or "Strip away assumptions"

Socratic Method (Advanced)

Use questions to guide discovery

• When to use: Learning, exploration, uncovering blind spots
• Why it works: Promotes deeper understanding through inquiry
• Example: "What questions should I be asking about my pricing strategy?"
• Tip: Frame as "Help me explore..." or "What am I not considering?"

Context Engineering (Advanced)

Strategically allocate context vs instructions for optimal AI reasoning

• When to use: Complex tasks where AI needs background knowledge, not just steps to follow
• Why it works: AI performs better with rich context than with overly detailed instructions. Giving Claude the 'why' behind a task enables better judgement calls.
• Example: "Here is our pricing strategy document and competitor analysis [uploaded files]. Based on this context, recommend pricing for our new service tier."
• Tip: Break complex tasks into focused chats instead of one mega-prompt. Use Claude Projects to provide persistent context. When results are poor, try adding more context rather than more instructions.

Bad → Better → Best Prompt (Beginner)

Transform vague prompts into high-quality instructions by adding context, files, and the AskUserQuestion pattern

• When to use: Every time you write a prompt — use this as a quality ladder to check your work
• Why it works: Each level adds structure that Claude uses to produce dramatically better output. The 'Best' level uses uploaded context files and asks Claude to interview you before starting.
• Example: Bad: "Write me a LinkedIn post about AI tools." Better: "I want to write a LinkedIn post about the AI tools I use daily. Don't start yet. Ask me clarifying questions first." Best: [Upload ABOUT-ME.md, ANTI-AI-STYLE.md, COPYWRITING.md] "I want to write a LinkedIn post about AI tools I use daily. First, read the uploaded files completely before responding. DO NOT start executing yet. Instead, ask me clarifying questions (use AskUserQuestion) so we can refine the approach together step by step."
• Tip: Always upload your About Me and Anti-AI Style files. End with "Ask me clarifying questions first" to trigger the AskUserQuestion pattern. Edit your prompt instead of sending follow-ups.

AskUserQuestion Pattern (Intermediate)

Let Claude prompt YOU instead of writing long instructions — clickable forms, real buttons, dramatically better results

• When to use: Complex tasks where you are not sure what Claude needs to know. Replaces long instructional prompts with a short request + interactive Q&A.
• Why it works: A 15-word prompt + clicking options costs almost nothing. A 500-word instruction gets re-read every time. Claude asks the right questions, you pick from options, and the output jumps 2-3 levels.
• Example: "I need a marketing plan for my plumbing business. Ask me clarifying questions first." Claude responds with: 1) What's your target area? 2) Monthly budget? 3) What services do you most want to promote? You click answers, Claude builds the plan.
• Tip: Use it for briefs you can't articulate. Works like a form for an employee. If the direction feels wrong, say "Develop 5 different strategies." You can even get a Mix & Match replacement: type journal and auto-paste your One Prompt. Settings.

Markdown Files as Prompt Libraries (Intermediate)

Replace traditional prompt libraries with .md files that Claude reads automatically — your voice, your rules, your examples as persistent context

• When to use: When you find yourself copy-pasting the same prompts or re-explaining your style, tone, and preferences in every conversation
• Why it works: Uploaded .md files become persistent context in Claude Projects and Cowork. Claude reads them before every response — so your brand voice, anti-AI style rules, and favourite prompt structures apply automatically without re-typing.
• Example: Create three files: about-me.md (who you are, what you do, how you think), anti-ai-style.md (banned words and patterns), copywriting.md (your writing rules and examples). Upload all three to a Claude Project. Every new chat in that project starts with full context.
• Tip: Keep each file under 200 lines. Upload once, reference forever. This is more powerful than copy-pasting prompts because Claude applies the rules across all your content automatically.