system_prompts_leaks

This repository collects and documents leaked system prompts for a wide array of popular AI chatbots and agents, including Anthropic's Claude variants, OpenAI's ChatGPT and Codex, Google's Gemini, and xAI's Grok, among others. It provides full, versioned system prompts dating back to May 2025, updated daily, making it a unique resource for understanding the hidden instructions that shape chatbot behaviour. The project is particularly useful for ML/AI engineers and researchers who want to analyse or replicate prompt engineering strategies used by leading models. With nearly 69,000 stars and recent commits as of October 2026, it shows strong community interest but no significant recent growth spikes. It is not suitable for teams seeking official or licensed prompt templates, or those wanting a tool for prompt generation or management rather than archival and analysis. For those needs, tools like LangChain or PromptLayer would be more appropriate.

Use this repository when you need access to comprehensive, versioned system prompts of major AI chatbots for research or reverse engineering. If you require officially supported prompts or prompt tuning frameworks, consider alternatives such as LangChain or PromptLayer instead, which focus on prompt construction and deployment rather than archival leaks.

This resource typically fits teams building AI research dashboards, analysing chatbot behaviour, or developing prompt-based evaluation tools. It is valuable for projects that require detailed insight into system prompt evolution across multiple AI providers, such as comparative studies or forensic prompt analysis.

Latest commit: 2026-10-03. Activity data is based on daily RepoPi snapshots of the GitHub repository.