How to use these prompts
Every prompt here is plain text you copy into an AI tool. The only extra step is filling in the blanks.
1. Find a prompt
Search the library or filter by category. Each card shows the model notes and the variables you'll need to fill in, so you can tell at a glance whether it fits your job. Click the title to open the full prompt.
2. Fill in the variables
Anything in {{double_braces}} is a variable: a slot for your own details. For example, Fact-check this claim has one variable, {{claim}}, and you replace it with the sentence you want checked.
- On a prompt's own page, type into the boxes under "Fill in the variables". The prompt text updates as you type, and Copy prompt copies the filled-in version.
- From a card in the library, the Copy button copies the prompt with the
{{variables}}still in it. Replace each one after you paste. - Paste the real thing. If a variable asks for an error, a diff or a document, paste the actual text, not a description or a screenshot of it.
- If a variable doesn't apply, write "none" or delete that line. Don't leave
{{braces}}in. The model will treat them as text and may guess what you meant. - Take out secrets first. Remove passwords, API keys and other people's personal details before you paste into any AI tool. For text that should stay on your machine, use a local model (see below).
3. Pick an AI tool
The prompts are plain text, so they work in any chat or coding tool. Each prompt's model notes tell you about any special needs: a vision-capable model, web browsing, long context, or a small local model.
| If you're doing this | Paste into |
|---|---|
| Everyday writing, research, marketing, explanations | A general chat assistant, such as ChatGPT, Claude or Gemini |
| Code that needs to see your project | A coding tool with access to your repo, such as Cursor or Claude Code |
| Fact-checks and anything recent | A tool with web browsing or search turned on |
| Long documents, logs or contracts | A model with a long context window |
| Private text, or working offline | A local model in Ollama or LM Studio (see the local LLM prompts) |
Prompts in the Agents category are system prompts and rules, not one-off questions. Their model notes say where they go, for example in an AGENTS.md or CLAUDE.md file, or in your agent's system prompt setting.
4. Check the output
AI models make mistakes, especially with facts, numbers, sources and code. Read the result before you rely on it. Run and test code, check that any links or citations exist, and look up figures yourself. If the first answer is off, reply with what's wrong; that usually works better than starting over.