Prompt Engineering Cheat Sheet (One-Page Reference)
A one-page, scannable prompt engineering reference: the five-part pattern, the highest-leverage fixes, and the mistakes to check for — no fluff, bookmark and reuse.
The condensed version. Five parts, one common mistake to check for first, three per-model notes, and a quick call on when to skip extra structure entirely. Bookmark this page instead of re-reading the full guides every time.
The five-part pattern
- Role — who the model should act as. Specific, not generic (“senior tax attorney,” not “expert”).
- Context — what a competent freelancer would need to know: audience, situation, what's already been tried.
- Task — an action verb and a deliverable, not a topic. “Compare X and Y on four dimensions,” not “tell me about X.”
- Constraints — length, tone, and explicitly what to exclude. Exclusions do more work than inclusions.
- Format — table, list, JSON, word count. The highest-leverage line in most prompts and the one most often skipped.
Check this mistake first
Before anything else: does your prompt state a task or just a topic? “Marketing strategy” is a topic. “Recommend one positioning angle for a $40/month SaaS tool, with a one-sentence reason” is a task. This single fix accounts for more improvement than any other item on this page.
Per-model notes
- Claude — follows format constraints the most literally. Be exact about structure; you'll get exactly that structure back.
- ChatGPT — the most forgiving of loose prompts, which makes it easy to underestimate how much structure would still help.
- Gemini — defaults to longer, more hedged answers. A hard length limit fixes this instantly.
When to skip the extra structure
Not every prompt needs the full pattern. Skip role and examples for well-known, single-step tasks (“summarize this email”). Skip “think step by step” on anything that isn't multi-step reasoning — modern models already do this internally when it matters. Full structure earns its cost on tasks where the default, generic answer wouldn't actually serve you.
Apply it right now
Paste a rough prompt and see the structured version, free, no login:
Go deeper
- The full five-part pattern, with per-tool sections and worked examples.
- Best practices and mistakes, expanded with the reasoning behind each.
- 12 before-and-after examples across writing, coding, analysis, and more.
- Named frameworks like CO-STAR and RACE, mapped back to this same pattern.
- Is prompt engineering still important? What changed and what didn't.
Frequently asked questions
What should be on a prompt engineering cheat sheet?
Five things, in order of leverage: the five-part prompt pattern (role, context, task, constraints, format), the single most common mistake to check for (topic instead of task), the per-model notes that actually change your approach, a short list of when NOT to bother with extra structure, and one link to try the pattern live. Anything longer stops being a quick reference.
Is a prompt engineering cheat sheet different from a full guide?
Yes — a cheat sheet is meant to be scanned in under a minute as a reminder, not read start to finish. It trades explanation for density. If you want the reasoning behind each item, the full guides are linked throughout this page.
Stop rewriting prompts. Try the one-click enhancer.
Try the PromptAI demoWorks where you do: the AI prompt enhancer everywhere, ChatGPT prompt enhancer in the browser, or the prompt enhancer for Claude Code in your terminal.