★ What Makes a Good Prompt
A great prompt is clear, specific, and gives the AI enough context to produce a useful response. The best prompts share these five qualities:
- Clarity — State exactly what you want. Avoid ambiguous or vague language.
- Specificity — Include details: audience, tone, length, format, constraints.
- Context — Provide background so the AI understands the situation and stakes.
- Structure — Use formatting (bullet points, numbered steps, sections) to guide the output.
- Iterability — Write prompts you can refine. Start simple, then add constraints based on the first result.
⚙ The Basic Formula
Almost every effective prompt can be built from four building blocks. Combine them and you get a reliable, repeatable structure:
Breaking it down
- ROLE — Tell the AI who to act as. Example: “You are a senior marketing copywriter.”
- TASK — State the action. Example: “Write a cold outreach email.”
- CONTEXT — Give background. Example: “The recipient is a VP of Sales at a SaaS startup with 50 employees.”
- FORMAT — Specify output shape. Example: “Return 150 words max, with a subject line and a single call-to-action.”
Put together, a complete prompt might read: “You are a senior marketing copywriter. Write a cold outreach email to a VP of Sales at a 50-person SaaS startup. The product is an AI-powered CRM. Keep it under 150 words, include a subject line, and end with one clear call-to-action.”
✎ 20+ Ready-to-Use Prompt Templates
Copy, paste, and replace the [bracketed] placeholders with your own details.
🌡 Temperature Guide
Temperature controls how random or focused the AI's output is. Lower = more deterministic. Higher = more creative. Pick the right setting for your task:
| Temperature | Behavior | Best For |
|---|---|---|
| 0.0 – 0.2 | Highly focused, deterministic, repetitive. The AI picks the most likely next token every time. | Code generation, data extraction, factual Q&A, formatting tasks, classification, math, legal/medical summaries where accuracy matters more than flair. |
| 0.3 – 0.5 | Slightly varied but still reliable. Mostly factual with minor creative flexibility. | Technical writing, documentation, email drafts, structured outlines, translations, summarization with consistent structure. |
| 0.6 – 0.8 | Balanced. Natural-sounding prose with reasonable variety without going off the rails. | Blog posts, marketing copy, articles, social media content, customer support replies, creative-but-grounded writing. |
| 0.9 – 1.0 | Highly creative, unpredictable, diverse. The AI explores less likely word choices. | Brainstorming, fiction/poetry, ideation sessions, creative naming, ad copy variations, out-of-the-box problem solving. |
Rule of thumb: If wrong answers are costly, go low. If boring answers are the problem, go high.
🧠 Advanced Techniques Advanced
✅ Do's and Don'ts
✅ Do
- Be specific about format, length, tone, and audience.
- Use examples (few-shot) when the task is non-obvious.
- Iterate — refine your prompt based on the first output.
- Assign a role to focus the AI's expertise and vocabulary.
- Break complex tasks into chained, smaller prompts.
- Use delimiters to separate instructions from data.
- Tell the AI what to do, not just what not to do.
- Specify constraints: word count, banned words, required sections.
❌ Don't
- Don't write vague prompts like “write something good.”
- Don't overload one prompt with too many unrelated tasks.
- Don't assume the AI knows your context — spell it out.
- Don't use contradictory instructions (e.g., “be brief” + “be thorough”).
- Don't forget to specify the output format (table, list, paragraph).
- Don't paste sensitive or confidential data into public AI tools.
- Don't trust factual outputs without verification — AI can hallucinate.
- Don't use high temperature for factual or analytical tasks.
📊 Quick Reference Table
| Task Type | Recommended Temp | Key Technique | Tip |
|---|---|---|---|
| Code Generation | 0.0 – 0.2 | Zero-shot + constraints | Specify language, framework, and coding style explicitly. |
| Summarization | 0.3 – 0.5 | Delimiters | Use ### tags around source text to avoid confusion. |
| Creative Writing | 0.8 – 1.0 | Role prompting | Assign a genre-specific persona for authentic voice. |
| Brainstorming | 0.9 – 1.0 | Few-shot + quantity ask | Ask for more ideas than you need, then curate. |
| Math / Logic | 0.0 – 0.2 | Chain-of-thought | Always ask to show reasoning step by step. |
| Classification | 0.0 – 0.3 | Few-shot | Provide 3–5 labeled examples covering each category. |
| Email / Communication | 0.4 – 0.6 | Role + format | Specify tone, length, and desired response/action. |
| Translation | 0.3 – 0.5 | Role prompting | Ask for cultural adaptation, not literal word-for-word. |
| Data Analysis | 0.2 – 0.4 | Structured output | Request tables and plain-language explanations. |
| Marketing Copy | 0.7 – 0.9 | Few-shot + role | Feed 2–3 examples of your brand voice first. |
| Research / Q&A | 0.0 – 0.3 | CoT + delimiters | Always verify cited facts independently. |
| Long-Form Article | 0.5 – 0.7 | Prompt chaining | Outline → draft → edit in separate prompts. |
📥 Download & Save This Cheat Sheet
Want this cheat sheet offline? Here's how to save it for quick access anytime:
- Press Ctrl + S (Windows/Linux) or Cmd + S (Mac) on your keyboard.
- Choose “Web Page, Complete” or “Page Source” as the save type.
- Pick a folder you'll remember — e.g., a “Reference” or “AI Tools” folder.
- Open the saved
.htmlfile in any browser — no internet required. - Bookmark this page (Ctrl + D / Cmd + D) for quick online access too.
You can also print a clean copy: press Ctrl + P / Cmd + P and choose “Save as PDF” for a portable, searchable reference.