Security audit
Security checks for vulnerabilities and agentic risk
This is a simple web-browsing guidance skill whose behavior is disclosed and purpose-aligned, with usability and privacy caveats rather than evidence of malicious activity.
Install only if you want an assistant that proactively searches the web, prefers Chinese-language and China-oriented sources, and summarizes in Chinese. Avoid putting sensitive private information into web queries unless you are comfortable sending those terms to search engines or websites.
The workflow instructs the agent to summarize results in Chinese, and the output section further requires using Chinese. This is a language policy constraint imposed unconditionally, with no opt-in or alternative for users who may prefer another language.
The instruction “使用中文” requires a specific language for all user-facing output. Because the file does not offer a language choice or justify a region-specific limitation, this is a natural-language policy violation.
The skill description says it can proactively search and browse the internet, but it does not warn users that their prompts may trigger requests to external websites and services. This reduces transparency and informed consent, especially when queries may be sent to search engines, content hosts, or browser automation targets.
The trigger conditions are very broad and match common requests like '了解', '学习', '查询', or '汇报', which can cause the skill to activate for routine conversations and initiate external network access without clear user intent. In a web-enabled skill, overbroad activation increases the risk of unnecessary browsing, privacy leakage through third-party queries, and unexpected tool use.
The skill hard-codes country: "CN" and search_lang: "zh" as default search parameters without indicating they are optional or user-driven. This can silently bias search results toward a specific jurisdiction and language, reducing result diversity and potentially exposing users to region-specific filtering, censorship, or relevance issues when they did not request such constraints.
No suspicious patterns detected.
