Security audit
Security checks for vulnerabilities and agentic risk
This shopping skill mostly does what it claims, but it sends queries and pasted shopping links to an external service while steering assistants away from clearly labeling affiliate-style redirects.
Review before installing if you are sensitive about shopping interests, pasted product links, referral tokens, IP address logging, or affiliate redirects. Use it only for explicit shopping/coupon tasks, avoid pasting private account-specific URLs, and treat '领券购买' links as external redirect links that may open platform apps or bind coupons to your account.
The README advertises extremely broad trigger phrases such as general shopping, recommendation, and price-related language that can cause the skill to activate for many unrelated user requests. Over-broad invocation increases the chance of unintended data disclosure to the external service and can hijack conversations where the user did not explicitly ask to use this shopping skill.
The privacy section states that the server records keywords, IP addresses, and timestamps for anti-abuse purposes, but this collection is presented in a reassuring way and not surfaced as a prominent operational warning before use. Because user queries may contain sensitive shopping interests or pasted links, logging this metadata without a clear just-in-time notice can create privacy risk and unexpected user data transmission to a third-party domain.
The trigger examples include broad purchase-intent phrases such as “我想买 xx”, “该不该买”, and “贵不贵”, which can match many ordinary shopping conversations that do not specifically request cross-platform price comparison or coupon aggregation. This can cause the skill to activate unexpectedly, sending user shopping-related queries to an external service and creating unnecessary data exposure and degraded routing quality.
The example issues a request to a third-party domain with the user's product query and then presents outbound affiliate-style purchase links, but the flow does not require an explicit disclosure or user consent before transmitting that query off-platform. This creates a privacy and transparency risk because user shopping intent and possibly sensitive search terms are shared with an external aggregator, and users may not realize links are mediated redirects.
The example forwards user-provided shopping links or tokens directly to an external domain, which can expose personal or sensitive data embedded in URLs such as referral identifiers, tracking parameters, account-related query strings, or proprietary coupon/share codes. In this skill’s context, users are encouraged to paste links and shopping tokens, so the privacy risk is real and amplified by the lack of any disclosure, consent step, minimization, or guidance about what data is transmitted.
The trigger table uses broad, common phrases such as “酒店”, “团购”, and “看电影”, which can cause the skill to activate in conversations where the user did not intend a coupon/comparison workflow. In an agent setting, over-broad invocation increases the chance of unsolicited external-link suggestions and data flows to the skill backend.
The rendered output encourages users to click coupon links that 'automatically open the app' and place coupons into their account, but it does not clearly warn about redirects, account binding, tracking, or platform-side effects. This can mislead users into taking actions with account and privacy consequences through an intermediary domain.
No suspicious patterns detected.
