Adversarial Testing & Prompt Injection Library
An empirical collection of zero-shot and few-shot jailbreak prompts, token-boundary anomalies, reasoning-trace hijacks, and leaked system instructions observed during local and frontier model evaluations.
Responsible AI Security Research & Defensive Alignment
All prompts and artifacts archived in this documentation are compiled solely for academic red-teaming, model alignment hardening, and defensive verification under the OWASP Top 10 for LLMs framework. Do not use for malicious activities.