Tests a model's resilience against adversarial prompts when operating under a system prompt that establishes an educational context for young students. This suite evaluates whether a 'tutor' persona can be exploited to bypass safety filters for incorrect, harmful, or dangerous content.