web-access

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly instructs preserving and passing full URLs that may contain session tokens/parameters (and shows curl examples embedding those URLs in request bodies), which requires the LLM to emit those token-containing URLs verbatim in generated commands—creating an exfiltration risk.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). 该 skill 的运行时 LLM 上下文会包含“页面/DOM 提取文本”，而这些文本来自用户日常浏览器通过 CDP 打开的网页内容（外部站点/用户未选择引入的内容），例如通过 /new→Page.navigate→/eval 读取 document.title/DOM 文本进入上下文，属于公共网页/第三方站点内容注入风险。