This project aims to expose and collect the system prompts used by major AI models and agents, enabling users to understand the underlying instructions that shape AI behavior and output. It targets AI researchers, developers, and power users who prioritize transparency and want to critically evaluate AI responses.

How It Works

CL4R1T4S operates as a community-driven repository for system prompts. It relies on user contributions through pull requests, encouraging the extraction, leakage, or reverse-engineering of these critical AI configuration files. The goal is to create a centralized, accessible database of these "hidden instructions."

Highlighted Details

• Collects system prompts for ChatGPT, Gemini, Grok, Claude, Perplexity, Cursor, Devin, Replit, and more.
• Aims to reveal AI limitations, forced personas, refusal behaviors, and embedded ethical/political frames.
• Encourages community contributions for model name/version, extraction date, and context.

Maintenance & Community

The project is maintained by @elder_plinius on X and Discord. Contributions are actively sought via pull requests.

Licensing & Compatibility

The repository does not explicitly state a license. Users should exercise caution regarding the use and distribution of collected prompts, as their original licensing is not specified.

Limitations & Caveats

The accuracy and completeness of the collected prompts depend entirely on community contributions. There is no guarantee that all prompts are up-to-date or accurately reflect the current behavior of the respective AI models.