leaked-system-prompts

Why we included this project

System prompts from widely used LLM-based services keep leaking into the open, and this archive collects them in one place, which makes it a useful window into how production chatbots are actually configured. Prompt engineers and developers can study how real products set tone, apply guardrails, and wire up tools in ways that typical tutorials never show. Researchers and students use it as a reference for comparing how different services frame model behavior, and the project is cited in academic papers. Every entry has to come with a verifiable source or a reproducible prompt, which keeps the collection far more reliable than a random pile of screenshots. It is a learning and reverse-engineering resource, not something you would deploy.