Security audit
Security checks for vulnerabilities and agentic risk
This is a straightforward web-search skill that discloses its Bocha API use, with privacy cautions around broad search activation but no evidence of hidden, persistent, or destructive behavior.
Before installing, understand that search terms derived from user prompts will be sent to Bocha's external API using your BOCHA_API_KEY. Use concise, non-sensitive queries and avoid searching secrets, credentials, private documents, or confidential business details unless you intend to disclose that content to the provider.
The skill explains that it performs web searches but does not clearly warn that user query content will be sent to an external provider. Without an explicit disclosure, users or higher-level agents may unknowingly transmit sensitive prompt contents off-platform, creating a privacy and data handling risk.
The skill instructs the agent to use web search whenever information is missing and even says to search when uncertain, which can cause unnecessary transmission of user queries to a third-party service. In contexts where user prompts may contain sensitive, personal, or proprietary information, this broad activation guidance increases privacy and data exposure risk even though the skill appears intended for normal search functionality.
The skill is explicitly designed to send requests to an external endpoint, which is expected for a web search tool, but still constitutes a real data egress boundary. Because the transmitted payload contains user-derived queries, any sensitive information included in those queries may be exposed to the third-party API if not minimized or filtered.
**请求端点 (Endpoint)**:
POST `https://api.bocha.cn/v1/web-search`
**请求头 (Headers)**:
The instruction hard-codes the fallback reply as a specific Chinese phrase with an English gloss, rather than letting the response match the user's chosen language. This can violate language-choice policy because it forces a locale-specific response without explicit user opt-in.
No suspicious patterns detected.