80+ curated prompts for ChatGPT, Claude, Gemini and other LLMs — across 8 categories: writing, coding, marketing, data analysis, creative, productivity, education and roleplay. Search by keyword, filter by category or model, copy in one click. Variable placeholders let you customize before pasting. Free, runs in your browser.
This is a hand-curated library of 80+ AI prompts tested across ChatGPT, Claude and Gemini — organized into eight categories that cover the most common LLM use cases: writing, coding, marketing, data analysis, creative work, productivity, education and roleplay. Every prompt has been written with care: clear instructions, explicit output expectations, and where useful, variable placeholders so you can plug in your own topic, audience or context before sending.
Unlike "prompt generator" tools that produce mediocre output by recombining clichés, every prompt here is something a human craftsperson would actually want to use. The point isn't volume — it's that each prompt teaches you a pattern you can reuse and adapt for similar tasks.
Step 1 — Browse or search. Use the category chips to narrow down (Writing, Coding, etc.), the model filter to surface what works best on your preferred LLM, or the search field to find anything by keyword.
Step 2 — Open a prompt. Click any card. The full prompt appears with variable inputs (when applicable).
Step 3 — Fill in variables. Many prompts have placeholders like {TOPIC}, {AUDIENCE} or {TONE}. The inline form lets you replace them before copying. You can also "Copy raw" to get the prompt with placeholders intact (useful for re-use templates).
Step 4 — Paste into ChatGPT, Claude, Gemini, or any LLM. All prompts are model-agnostic, but some perform better on certain models. The badge on each card indicates the recommended model — usually Any.
As of 2026, the frontier closed-source models (GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro) are roughly equivalent in raw capability, but each has its strengths:
If a prompt is tagged Any, it works equally well on all three. Use the model filter chips at the top to focus on what you're using.
Many prompts include placeholders in curly braces, like {TOPIC}, {AUDIENCE} or {STYLE}. These are clearly marked. When you open a prompt card, the variable fields appear automatically. Fill them in, click Copy with variables, and the prompt is copied with your values substituted.
If you prefer to copy the template as-is and customize it manually in your LLM chat, use Copy raw.
AI prompt generators (tools that ask the LLM to generate a prompt for another task) produce vague, formulaic output that includes every cliché in the training data: "You are a world-class expert in...", "Please provide a detailed, comprehensive analysis..." etc. The result is verbose, repetitive, and produces middle-of-the-road answers.
A hand-crafted prompt — built by someone who has actually iterated on the output — is shorter, more specific, and gives the model a clear task. Every prompt in this library has been tested. None of them include the phrase "world-class expert".
Explain the following code line by line. For each line, describe what it does in plain English, and flag any subtle behaviors or potential bugs.
```
{CODE}
```
After the breakdown, give a 2-sentence summary of what the entire snippet does.Write 5 different headline variations for {PRODUCT_NAME}, targeting {AUDIENCE}.
Each headline should test a different angle:
1. Benefit-focused
2. Pain-point focused
3. Question-based
4. Curiosity gap
5. Direct + specific number
After the list, briefly explain which one you think performs best and why.Rewrite the following text in a {TONE} tone, keeping the same core meaning and length.
Original:
{TEXT}
Output only the rewritten text — no explanation, no preamble.A prompt is the instruction you give to a large language model. The quality of your prompt directly determines the quality of the response. Good prompts are specific, give the model a clear task, define the desired output format, and provide context when needed. Bad prompts are vague or open-ended.
Prompt engineering is the craft of designing prompts that produce reliable, high-quality outputs from LLMs. Core techniques: be specific about the task, define the output format, give examples (few-shot), break complex tasks into steps (chain-of-thought), and assign a role when useful. This library is a starting point — adapt the prompts to your context.
Every prompt here is one I've personally used or tested — for writing, coding, marketing, or productivity work. The library is intentionally small and focused. It includes the patterns that produce reliably good output across all major LLMs, not every prompt that exists.
Mostly yes. Modern frontier models (GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro) follow well-structured prompts equally well. Differences emerge in nuance: Claude tends to write longer and more thoughtful prose; GPT is faster and more direct; Gemini handles huge documents best. The model badge on each card indicates the best fit when there's a meaningful difference.
Absolutely — these are templates, not commandments. Use them as starting points and adapt the wording, instructions, output format and constraints to fit your specific use case. The point of a curated library is to give you patterns you can build on, not to make you copy-paste exact text.
No — feel free to use, modify, redistribute, or build on any prompt here without attribution. Short generic instructions like these are not copyrightable in most jurisdictions, and the goal of the library is to spread good prompting patterns.
Click the ♥ icon on any card to favorite it. Favorites are stored in your browser's localStorage (no account needed). Use the "♥ Favorites" button at the top of the toolbar to filter the library down to your saved prompts.
Yes. Everything runs in your browser. No prompts you fill out or copy are sent to any server. Your favorites are saved locally on your device only. No signup, no tracking of prompt content.