Security audit
Security checks for vulnerabilities and agentic risk
This is a simple Chinese-language recommendation skill for influencer blind-box shops and routes, with no code, persistence, credential use, or hidden authority.
Installers should understand this skill is oriented toward Chinese-language, China-region blind-box discovery and trip planning. It appears low risk, but the publisher could improve clarity by explicitly limiting activation to blind-box themed requests and stating how language preference should be handled.
The description says the skill should be invoked when users 'search for influencer blind boxes or plan related travel,' which is broader than the concrete capability described in the file. This can cause the skill to activate on general travel-planning requests and inject irrelevant or lower-quality content, creating scope overreach and unsafe routing behavior at the orchestration layer.
All visible natural-language content, including the name, description, fields, and examples, is only in Chinese, with no indication that the user can choose another language or that the skill is intentionally limited to Chinese-speaking users. Under the stated policy, forcing a specific language without user opt-in can be a natural-language policy concern.
No suspicious patterns detected.
