Security audit
Security checks for vulnerabilities and agentic risk
This skill is an AppGrowing ad-analysis integration, but users should review it because it can send prompts to an external service and accepts API keys pasted into chat.
Install only if you intend to use AppGrowing/YouCloud for ad creative analysis. Prefer configuring YOUCLOUD_API_KEY securely as an environment variable, do not paste API keys or confidential business data into chat, and expect analysis prompts to be sent to the external AppGrowing/YouCloud service.
The skill explicitly tells the agent to accept API keys pasted directly into chat and use them for the request, without any safety warning. Chat is commonly logged, retained, and visible to platform operators or other integrations, so encouraging credential submission through chat creates a direct risk of secret exposure and account compromise.
The trigger phrases include broad, common-language terms such as '分析素材' and '素材解析', which can plausibly appear in ordinary conversation and cause the skill to activate unintentionally. Because the skill forwards user input to an external API, unintended activation can lead to accidental disclosure of user prompts or context to a third party and unexpected use of privileged API-backed behavior.
The instruction to activate whenever '用户意图明显是触发 agclaw' uses vague intent matching instead of a strictly bounded trigger, increasing the chance of the model invoking the skill based on ambiguous context. In this skill, that ambiguity is more dangerous because invocation may cause user content to be transmitted to an external service without an explicit, fresh consent boundary.
The API specification describes sending the user's input to AppGrowing's external endpoint but does not clearly warn users in the skill's user-facing description that their prompts will be transmitted to a third-party service. This weakens informed consent and can expose sensitive business, marketing, or proprietary data if users assume analysis is local to the host assistant.
This markdown file uses Chinese throughout headings, prompts, and outputs, but does not indicate that the skill is China-specific or provide any user opt-in for language/locale. Under the policy, forcing a specific language without user choice is a natural-language policy violation.
No suspicious patterns detected.
