This repository contains prompt injection techniques and jailbreak attempts designed to manipulate AI language models into ignoring their safety guidelines and operational constraints. Developers working in AI security, red-teaming, or adversarial testing might use this to identify vulnerabilities in LLM systems and understand how models respond to adversarial inputs. However, this is primarily a collection of attack prompts rather than a legitimate development tool, and using it for anything other than authorized security research could cause harm.

76 open issues · last updated Jul 22, 2026