Security audit
Security checks for vulnerabilities and agentic risk
This is a low-risk shopping advice skill with one off-topic business-planning example that should be cleaned up but does not create malicious or high-impact behavior.
Before installing, note that this skill is appropriate for shopper-facing merchant trust decisions on Chinese ecommerce platforms. The publisher should remove or split out the night-market food-stall example because it broadens the skill into small-business advice, but the current package is Markdown-only and does not ask for sensitive access.
The file centers the skill on Chinese ecommerce platforms and all concrete user examples are in Chinese, but it does not explicitly say users may interact in another language or choose their preferred locale. That can amount to an implicit locale constraint without opt-in.
The skill includes an unrelated scenario about starting a night-market food stall, which materially expands behavior beyond ecommerce merchant-trust evaluation. This can cause an agent to provide business-operational and food-sales guidance in contexts where the skill is expected to stay narrowly scoped, increasing the chance of unsafe or policy-inconsistent outputs.
The document says the skill is 'not a seller onboarding plan' but then provides concrete seller-startup advice for operating a night-market food stall, including product selection and sourcing. This contradiction weakens scope boundaries and may lead an agent to ignore stated constraints, producing off-domain operational guidance with legal and safety implications.
An unrelated usage scenario broadens the apparent scope from purchase-risk evaluation to entrepreneurial stall-planning advice. Scope drift is dangerous because agents often generalize from examples, so one off-topic scenario can override the intended boundaries of the skill and trigger responses in unintended domains.
No suspicious patterns detected.
