Security audit
Security checks for vulnerabilities and agentic risk
This skill is mostly a coherent ecommerce page-design workflow, but it includes ungated watermark-removal instructions for commercial product images and broad triggers that could apply it outside its niche.
Review before installing. The skill does not show malware-like behavior, persistence, credential access, or remote code execution, but it should only be used on product images the user owns or is authorized to modify. Its activation criteria should also be narrowed so ordinary product-page design requests do not automatically inherit TCM health-claims guidance or watermark-removal behavior.
The skill provides concrete watermark-removal instructions but gives no warning to verify ownership, authorization, licensing, or legal restrictions before doing so. This is dangerous because watermark removal can facilitate copyright infringement, provenance concealment, or misuse of third-party product images in a commercial setting.
The workflow explicitly includes removing watermarks and then exporting clean commercial assets, which operationalizes potentially infringing image laundering as part of normal use. In ecommerce context this is especially risky because it can enable unauthorized reuse of competitors' or licensors' product imagery at scale, increasing legal and trust harms.
The trigger phrases in the metadata are broad enough to match generic requests like '做详情页' or '产品详情页', which can cause the skill to activate outside its intended niche. Over-broad activation is dangerous because it can unexpectedly steer unrelated user tasks into a workflow that includes compliance-sensitive claims and watermark-removal steps.
The '何时使用' section lists ambiguous activation cues without boundaries, so the agent may invoke the skill for ordinary design help that merely mentions detail pages or export slices. In this skill, accidental activation is more dangerous because later steps include regulated-health marketing guidance and image-manipulation instructions that should not be applied broadly.
No suspicious patterns detected.
