Inspired by elder-plinius/CL4R1T4S.
Source: elder-plinius/CL4R1T4SReverse-engineered from real GitHub workflow.
LEAKED SYSTEM PROMPTS FOR CHATGPT, GEMINI, GROK, CLAUDE, PERPLEXITY, CURSOR, DEVIN, REPLIT, AND MORE! - AI SYSTEMS TRANSPARENCY FOR ALL! 👐
No specific use case defined.
Hey, I'd like to set up a new open-source project similar to CL4R1T4S. The core idea is to create a public repository dedicated to fostering transparency in AI by collecting and sharing "leaked" or reverse-engineered system prompts, guidelines, and operational instructions from various major AI models and agents. This project aims to expose the hidden inputs that shape AI behavior, helping users understand what AIs are told to do, what they can't say, and the ethical frames they operate within. The goal is to make the unseen inputs visible, because "to trust the output, one must understand the input."
For the architecture, it's pretty straightforward: I envision a GitHub repository acting as the central hub. The primary "tech stack" here is really just a structured collection of text-based files managed with Git for version control and public access. We're not building a complex application, but rather a curated data archive.
Could you help me set up the initial structure for this?
1. **Root Level:**
* A `README.md` file that clearly articulates the project's mission, explaining *why* this transparency is crucial. It should also outline how contributors can submit new prompts, specifying required information like model name/version, extraction date, and any relevant context.
* A `LICENSE` file (you can pick a common open-source one like MIT for now).
2. **Core Content Structure:**
* Create separate top-level directories for major AI providers. Based on the CL4R1T4S project, examples include `OPENAI`, `GOOGLE`, `ANTHROPIC`, `XAI`, `PERPLEXITY`, `CURSOR`, `DEVIN`, `REPLIT`, `META`, `MISTRAL`, and `VERCEL V0`.
* Inside each of these provider directories, generate a few placeholder text files (e.g., `model_xyz_prompt.txt`, `guidelines_v1.md`) that would eventually contain the actual system prompts or operational instructions. These placeholders should include comments indicating where the model name, version, and extraction date would go to guide future contributions.
The simplicity of this file-based structure and relying on GitHub for version control and distribution is key, making it highly accessible for both contributions and consumption, without needing a complex backend or frontend. It's essentially a transparent data repository for AI system intelligence.