Skip to main content

System Prompts Leaks

Welcome to the most comprehensive collection of system prompts extracted from popular AI chatbots including Claude, ChatGPT, Gemini, Grok, and many others.

What is This Collection?

This repository contains extracted system prompts (also known as system messages or developer messages) that define how AI chatbots behave, respond, and interact with users. These prompts are the foundational instructions that shape the personality, capabilities, and limitations of modern AI assistants.

Anthropic Claude

System prompts for Claude models including Opus, Sonnet, and specialized versions

OpenAI ChatGPT

Complete collection of GPT-5, GPT-4, and o-series model prompts

Google Gemini

Gemini 3, Gemini 2.5, and workspace-specific system prompts

xAI Grok

Grok 4.2, 4.1, 4, and personality variations

Why System Prompts Matter

Understanding system prompts helps you:
  • Learn prompt engineering - See how leading AI companies craft instructions
  • Understand AI behavior - Discover what shapes AI responses and limitations
  • Research AI safety - Study how constraints and guidelines are implemented
  • Improve your prompts - Apply techniques used by professional prompt engineers
This collection is maintained for educational and research purposes. All prompts were extracted through documented prompt injection techniques and are publicly shared for transparency.

What’s Included

Our collection spans multiple AI platforms and includes:
  • Current production system prompts for major AI models
  • Tool-specific prompts (Canvas, Memory, Search, Computer Use, etc.)
  • API vs. web interface variations
  • Personality and persona variations
  • Historical/archived versions showing prompt evolution
  • Platform-specific implementations (Chrome extensions, desktop apps, Excel add-ins)

Quick Navigation

What Are System Prompts?

Learn the fundamentals

How to Use This Collection

Get started with the prompts

Contributing

Add new prompts to the collection
Comprehensive system prompt including computer use capabilities, skills framework, artifact creation, persistent storage, and MCP server integration.
Complete system message covering personalization, content references, image groups, tool usage, and response behaviors.
Full system instructions including execution steps, safety guidelines, tool usage rules, and response formatting.
System prompts showing how Grok implements different personalities and safety constraints.

Stay Updated

This collection is continuously updated as new models are released and system prompts evolve. Star the GitHub repository to stay informed about updates.
Have a system prompt to contribute? Check out our contributing guide to add it to the collection.

Build docs developers (and LLMs) love