use-gemini

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

• Third-party content exposure detected (high risk: 0.85). 该技能的 web-search 工具在运行时会抓取/汇总公共网页搜索结果并把其文本（工具返回的 content[0].text）注入到代理的 LLM 上下文，因此存在来自“公共 web content fetched at runtime”的外部自由文本风险。

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill requires running the MCP server via "npx -y @tuannvm/gemini-mcp-server", which fetches and executes remote npm package code at runtime and that server provides tools (e.g., shell) that can control prompts and execute commands, so this external dependency is used at runtime and is required.