Security audit
Security checks for vulnerabilities and agentic risk
The skill is a coherent product-authenticity checker, but it can save receipts, photos, and screenshots locally and use external lookup sites without clear opt-in or retention limits.
Install only if you are comfortable with product details, company names, links, and image-derived information being queried on external regulatory or business sites, and with local evidence files such as receipts, product photos, and screenshots being saved. Avoid sharing unrelated personal information in images, redact receipts where possible, and delete generated evidence packages when no longer needed.
Automatically treating any product image as consent to start the skill is unsafe because images may contain receipts, addresses, chat screenshots, or other sensitive data unrelated to certification checks. In this skill's context, auto-activation can trigger external queries and screenshot retention without an explicit opt-in, amplifying privacy exposure.
All instructions, triggers, outputs, and examples are presented exclusively in Chinese, and the workflow assumes Chinese-language interaction and Chinese product-regulatory context without explicitly offering language opt-in. This can amount to a language policy violation unless the locale constraint is clearly documented and justified for users.
The invocation phrases are very broad and match ordinary requests like 'check this product' or 'see if this is legitimate,' which can cause the skill to activate in situations where the user did not specifically consent to an investigative workflow. In this skill, activation can lead to data collection, external lookups, and evidence retention, so overbroad routing meaningfully increases privacy and consent risk.
The skill directs searches across government and third-party sites using product, company, and possibly user-supplied purchase details, but it does not disclose that this information may be transmitted to external services. This creates a privacy and data-sharing risk, particularly when using commercial platforms like Tianyancha or QCC that may log queries and correlate them with the session.
The skill instructs immediate screenshot capture and local preservation of query-result pages, but it does not clearly warn the user that these artifacts will be stored on disk. Because the workflow also references receipts and packaging photos, silent local retention can expose personal and purchase data beyond the user's expectations.
The evidence-package design explicitly persists product photos, receipts, and screenshots in a local workspace directory, creating a durable store of sensitive consumer evidence. If the device, workspace, or other skills are compromised, these files could reveal identities, purchases, locations, and dispute activity, making the risk greater than transient processing.
