jailbreaks — Curated collection of LLM jailbreak prompts and bypass techniques, documenting adversarial inputs that circumvent AI model safety guardrails.