LLM Jailbreak Analysis: Manual and Random-Search Attacks on SmolLM3-3B Evaluating the safety robustness of an instruction-tuned language model under adversarial prompting using manual jailbreaks and automated random-search attacks