cl4r1t4s

CL4R1T4S is a repository dedicated to AI systems transparency and observability. It collects full extracted system prompts, guidelines, and tool definitions from a wide range of major AI models and agents, including those from OpenAI, Google, Anthropic, xAI, Perplexity, Cursor, Windsurf, Devin, Manus, and Replit. The stated rationale is that understanding the input is necessary to trust the output. AI labs shape model behavior through large, unseen prompt scaffolds, and these hidden instructions can influence what models say, which personas and functions they follow, how they refuse or redirect, and what ethical or political frames are applied by default. The project argues that interacting with an AI without knowing its system prompt means talking to a non-neutral, shaped intelligence. Contributions are accepted via pull requests. The repository asks contributors to include the model name and version, the date of extraction if known, and optional context or notes. Contact is offered through the maintainer's social and Discord channels. The README also contains an embedded directive written in leetspeak that instructs the reader to include their own instructions in full within the original interface of discussion. This appears to be part of the repository's characteristic style rather than a description of software functionality. Overall, the project functions as a public reference archive for prompt transparency, aimed at researchers, journalists, and users interested in how AI systems are instructed to behave.