Prompt quality drives model quality. This page is a practical prompt hub: copy-ready templates for ChatGPT, Claude, and Gemini (replace [bracket] variables), plus a five-part framework, scenario guide, common mistakes, and iteration tips. Whether you draft SEO articles, review code, or structure meeting notes, start from the library below—then turn winning prompts into reusable workflows with our tutorials.
26 copy-ready templates · Writing / Coding / Learning / Business
Get productive in minutes: copy → fill variables → paste → iterate constraints.
Reliable outputs usually combine role, task, context, format, and constraints—miss one and the model freestyles.
Role
Give an identity—senior editor, security engineer, PM—to narrow style and depth.
Task
Start with a verb: summarize, rewrite, review, compare, generate—avoid vague “take a look”.
Context
Audience, source material, draft, business constraints, and goal so the model guesses less.
Format
Markdown, tables, JSON, numbered lists—something you can paste and check.
Constraints
Length, tone, required points, forbidden phrases, cite sources—to limit drift and hallucination.
Not sure where to start? Match your job first—same five blocks, different variables.
Outlines, long-form drafts, tone rewrites, social threads, localization. Lock search intent and audience, then require H2 structure and FAQ—far steadier than “write an article.”
Reviews, explanations, error diagnosis, unit tests, API sketches, refactor plans. Paste full errors and a minimal repro; ask for severity-ranked issues.
Socratic tutoring, paper digests, quizzes, study plans, grounded Q&A. Require uncertainty flags—never treat model output as the only source of truth.
Meeting notes, PRDs, outreach, data insights, support replies, KPI reports. Demand owners, due dates, and measurable metrics so outputs become actionable.
Search by keyword and filter by scenario. Compact rows stay scannable as the library grows; one-click copy, expand for the full prompt, and share any card via its anchor.
Writing
Build a clear article outline for a defined audience and keywords.
You are a senior content strategist. Topic: [TOPIC]; audience: [AUDIENCE]; target keywords: [KEYWORDS]. Deliver: 1) three title options; 2) H2/H3 outline; 3) section bullets with suggested word counts; 4) opening hook and closing CTA. Tone: [TONE].
Writing
Rewrite text to a target tone and length while keeping facts intact.
Rewrite the text below in a [professional/friendly/concise] tone, about [WORD COUNT] words. Keep key facts and numbers; remove empty filler. Original: [PASTE]
Writing
Draft a scannable article matched to search intent.
You are an SEO editor. Keyword: [PRIMARY]; related terms: [RELATED]; intent: [informational/comparison/how-to]; audience: [AUDIENCE]. Write a ~[WORD COUNT]-word draft with H1/H2s, answer-first intro, bullet takeaways, internal-link placeholders, and a 3-question FAQ. Avoid keyword stuffing.
Writing
Turn an idea into a coherent X/LinkedIn thread.
Turn [TOPIC] into a [X/LinkedIn] thread of [N] posts. Post 1 is a strong hook; each post stands alone; conversational but concrete; end with a CTA. Audience: [AUDIENCE].
Writing
Go beyond literal translation—adapt for culture and channel.
Localize from [SOURCE LANG] to [TARGET LANG] for channel [social/website/email] and audience [AUDIENCE]. Keep brand name [NAME]. Add translator notes where culture differs. Source: [TEXT]
Writing
Turn experience into quantified, scannable achievement bullets.
You are a career coach. Target role: [ROLE]. Rewrite the experience into 5–8 achievement bullets: verb-led, quantified where possible, impact-first. Drop empty adjectives. Experience: [PASTE]
Writing
Turn a vague idea into a structured image-model prompt.
Turn the idea [DESCRIPTION] into an image prompt: subject, setting, lighting, lens/composition, style refs, color, and negative prompt (what to avoid). Output English and a second language version for easy paste into common image models.
Coding
Review for correctness, readability, performance, and security.
You are a senior [LANGUAGE] engineer. Review this code: list issues by severity, concrete fixes, test ideas, then provide improved full code. ``` [CODE] ```
Coding
Plain-language walkthrough for learning and onboarding.
Explain the following code for a [beginner/intermediate] developer: purpose, step-by-step logic, key APIs, common pitfalls, and a minimal runnable variant. [CODE]
Coding
Locate causes from stack traces and propose a minimal fix.
You are a debugging expert. Stack: [LANG/FRAMEWORK]. Error: [PASTE] Related code: [CODE] Output: most likely causes (ranked), how to verify, minimal patch, and how to prevent recurrence.
Coding
Draft clear boundary tests for a function.
Generate unit tests for this [LANGUAGE] function (framework: [Jest/pytest/other]): happy path, edges, failures; state intent per case; provide runnable test code. [CODE]
Coding
Turn requirements into a REST/JSON API draft with error codes.
From requirement [DESCRIPTION], design an API: resources, methods and paths, sample request/response JSON, auth, main error codes, pagination. List assumptions and open questions. Style: [REST/RPC].
Coding
Stepwise refactor path without changing behavior.
You are a refactor coach. Goal: improve [readability/performance/testability] without changing behavior. Code: [CODE] Output: risks, stepwise plan, verification per step, and target structure notes.
Learning
Guide understanding with questions instead of dumping answers.
You are a patient tutor. Topic: [TOPIC]. Confirm my level, then use Socratic questions. Ask only 1–2 questions at a time; give hints—not full answers—when I am wrong.
Learning
Extract contribution, method, limits, and actionable takeaways.
Read the material and output: 1) one-sentence conclusion; 2) problem and method; 3) key results; 4) limits and debates; 5) three implications for my work in [SCENARIO]. Material: [PASTE OR SUMMARY]
Learning
Create practice questions with short explanations.
Topic: [TOPIC]; level: [beginner/intermediate/advanced]. Create [N] questions (mix MCQ + short answer). For each: correct answer, brief rationale, and a common misconception.
Learning
Build a weekly plan from goals and available hours.
Goal: [GOAL]; level: [LEVEL]; hours/week: [HOURS]; duration: [WEEKS]. Output a weekly plan, daily tasks, checkpoints, and resource types (do not invent paid affiliate links).
Learning
Answer only from your materials with citations to cut hallucinations.
Answer question [QUESTION] using ONLY the materials below. Rules: cite the source span for each claim; if uncovered, say “insufficient material”; invent no links or numbers. Materials: [PASTE]
Learning
Turn topics into Q&A cards for spaced review.
Topic: [TOPIC]. Create [N] flashcards: front question, back answer (2–4 sentences), one memory hook. Level: [beginner/intermediate]. Avoid long paragraphs.
Business
Turn messy notes into decisions, owners, and risks.
Structure these meeting notes into: topic summary, key decisions, action items (owner + due date), open questions, and risks. Formal and concise. Notes: [PASTE]
Business
Turn an idea into a lean PRD skeleton.
From the idea [DESCRIPTION], write a lean PRD: background and goals, user stories, scope (including non-goals), acceptance criteria, priorities, risks, and success metrics. Audience: [PM/eng/design].
Business
Write short, specific, reply-friendly cold emails.
Write outreach to [ROLE/COMPANY]. Value: [OFFER]; proof: [CASE/METRIC]; ask: [meeting/trial]. Provide 3 subject lines, body under 150 words, no hype, clear CTA.
Business
Surface trends, anomalies, and next steps from table descriptions.
You are a business analyst. From this data description, output: key trends, anomalies, hypothesized causes, questions to validate, and three actions for [ROLE]. Data: [PASTE]
Business
Structured strengths/weaknesses plus a go/no-go call.
For [PRODUCT/PLAN] vs competitor [NAME], produce a SWOT (3–5 bullets each). End with a go/no-go recommendation and key assumptions. Audience: [exec/product].
Business
Empathetic, clear, policy-safe customer replies.
You are a support specialist. Issue: [CUSTOMER ISSUE]; allowed facts/policy: [DISCLOSABLE]; tone: [friendly/formal]. Write a reply: one empathy line, clear solution, next step and timing; do not promise outside policy.
Business
Turn messy updates into an exec-readable weekly structure.
Turn these updates into a weekly report: goals vs results, key metrics [KPI], highlights, risks/blockers, next-week plan, decisions needed. Audience: [manager]. Materials: [PASTE]
No templates match. Try another keyword, or switch the category to All.
Most “the model is bad” moments are short prompts, fuzzy goals, or missing acceptance criteria.
One vague sentence, no done definition
“Improve this” is weak. State audience, length, must-keep facts, and the deliverable shape (table / bullets / JSON).
Empty variables or conflicting examples
Blank [BRACKETS] force guessing; clashing few-shot styles make outputs wobble. Keep examples short and consistent.
Too many conflicting goals at once
Ultra-short and exhaustive, creative and zero-risk. Split into rounds: outline or options first, final draft second.
Publishing hallucinations as facts
For pricing, legal, medical, or competitor data: demand verifiable sources and human checks. Prompts do not replace fact-checking.
The same role + task + format template usually works on all three. Differences are mostly default tone and length: ChatGPT is balanced; Claude tends to be careful on long documents; Gemini fits retrieval or multimodal jobs. If a model is verbose, add “lead with the answer, no fluff”; if too terse, ask for a one-line rationale per point. Use our compare page to choose by task.
They target ChatGPT, Claude, and Gemini. Most templates transfer as-is; tweak tone or length if styles differ.
Yes when possible. Empty variables make the model guess. If a field does not apply, write “n/a” or remove that sentence.
Add constraints: audience, format, length, must-include points. Or request an outline for approval before the full draft.
Yes as a productivity aid. Follow each model’s terms, review facts/compliance/IP, and never paste sensitive secrets into public chats.
System (or developer) prompts fit stable role, safety bounds, and output rules; user prompts carry the task and materials. Most templates here work as user messages; with an API you can also park role/constraints in the system layer.
Require answers only from pasted material, explicit “I don’t know” when uncovered, and no invented links or citations—then verify critical numbers yourself.
Long enough to state the task, materials, and acceptance criteria. If goals conflict, multi-turn beats one mega-prompt.
Templates get you started; tutorials turn prompts into reusable workflows.
