CL4R1T4S
CL4R1T4S is an open-source GitHub repository that aggregates systematically leaked system prompts from major AI assistant platforms, including ChatGPT, Claude, Gemini, Grok, Perplexity, Cursor, Lovable, and Replit. The repository aims to provide transparency into the hidden initial instructions that govern the behavior of these AI systems. The repository contains text files or documentation that expose the system prompts used by these AI models, which are typically not disclosed by their providers. By collecting these leaked prompts, CL4R1T4S allows researchers, developers, and enthusiasts to study the underlying rules, constraints, and ethical guidelines embedded in commercial AI assistants. The project is maintained by the user elder-plinius and is presented as a resource for AI systems transparency. It highlights the importance of open access to information about how AI models are instructed and controlled, fostering discussions about accountability, bias, and the hidden design choices in AI products. Although the repository primarily serves as a dataset of leaked prompts, it has become a reference point for prompt engineering research and for understanding the differences and similarities between AI assistant configurations across various platforms.
- Category
- Prompt Engineering
- Pricing
- Open Source
Problem solved
CL4R1T4S addresses the lack of transparency in AI systems by exposing the system prompts that control their responses. This enables researchers and developers to understand underlying biases, restrictions, and design choices, contributing to more informed discussions about AI ethics and accountability.
Use cases
- Researching AI system prompt design
- Auditing AI model behavior and constraints
- Developing prompt engineering techniques
- Comparing system prompts across AI platforms
- Teaching about AI safety and transparency
- Creating custom system prompts for local models
Audience
- Developers
- ML Engineers
- Data Scientists
- Researchers
- Hobbyists
Pros
- Provides unprecedented transparency into AI system prompts
- Comprehensive coverage of major AI assistants
- Openly accessible for research and education
- Encourages discussion on AI ethics and accountability
- Easy to browse and compare prompts
Cons
- Contains leaked proprietary information with potential legal and ethical issues
- Prompts may be outdated or inaccurate
- No guarantee of authenticity
- May encourage misuse of system prompt knowledge
- Lacks curation or context for each leak
Official website: https://github.com/elder-plinius/CL4R1T4S
Canonical LambdaBase page: https://www.lambdabase.com/tools/cl4r1t4s