chatgpt_system_prompt is a collection of GPT system prompts and documentation of prompt injection and leaking techniques.
The project addresses the problem of understanding how GPT system prompts work and how they can be manipulated or extracted. It gathers actual system prompts used by various GPT implementations alongside techniques for prompt injection and knowledge leaking, providing researchers and developers with concrete examples of both the prompts themselves and methods to bypass or extract information from them.
This collection is most useful for security researchers, prompt engineers, and developers working on AI safety and robustness. It suits projects where understanding prompt vulnerabilities is important, such as building defenses against prompt injection attacks or studying how language models can be manipulated. The repository serves as a reference resource rather than a tool to be integrated into other projects.
The project maintains an active collection of examples and techniques, with contributions documenting newly discovered prompts and injection methods as they emerge in the field.
