Security audit
Security checks for vulnerabilities and agentic risk
This is a simple infant feeding and schedule planning skill with disclosed medical boundaries and no code, credential access, persistence, or data export behavior.
Install only if you want general baby food and schedule planning help. Do not rely on it for emergencies, medication or supplement dosing, significant allergies, premature infants, abnormal growth, choking events, or other medical-risk situations; use a qualified clinician for those cases.
The trigger keywords are broad, generic parenting phrases such as '宝宝作息', '过敏怎么办', and '几个月吃什么', which can match many ordinary conversations unrelated to intentional skill invocation. This can cause accidental activation and unsolicited medical-adjacent guidance, increasing the chance that users receive advice in contexts where the skill lacks sufficient context or should defer to a clinician.
The natural-language examples are open-ended and ambiguous, encouraging activation from loosely related requests without defining when the skill should abstain or hand off. In a maternal/infant feeding context, that ambiguity is more sensitive because the skill may respond to allergy, feeding, or sleep questions that can implicate health and safety decisions.
No suspicious patterns detected.
