Study LLM Prompts Across Platforms

What problem does this solve?

Gives you access to thousands of real system prompts, custom instructions, and security techniques across LLM platforms. Learning how other prompt engineers build instructions is difficult without seeing actual implementations that work. This repository removes the guesswork by collecting reverse-engineered prompts from ChatGPT, Claude, Gemini, and others, plus articles explaining prompt engineering and LLM security.

How does it work?

Browse or search the repository by LLM platform (ChatGPT, Claude, Gemini, Copilot, etc.) or by category (custom instructions, system prompts, jailbreaks, security protections). Each file contains a real prompt structure you can study, adapt, or use as reference. Articles cover reverse engineering techniques, ChatGPT sandbox contents, and prompt injection security.

Included Skill Groups

What's the biggest win?

Learn prompt engineering patterns from 1,600+ real examples instead of starting from zero.

What should I know technically?

All prompts are plain text or markdown, readable in any editor. System prompts may contain JSON, XML, or structured formats specific to their vendor. Large markdown files (1,573 ChatGPT entries) load quickly in version control.

What should I watch out for?

Many prompts reflect experimental or adversarial techniques. Some jailbreaks may be outdated as LLM guardrails evolve. Instructions are for learning only; using them to bypass intended model behavior violates most LLM terms of service. Some custom GPT entries reference deprecated functionality.

Key Features