geo-content-optimizer

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

• Third-party content exposure detected (high risk: 0.95). 该技能在 Phase 1/4 使用 mcp__web_reader__webReader 与 Playwright (mcp__playwright__browser_navigate/mcp__playwright__browser_snapshot) 读取用户提供的任意 URL 页面正文/标题，并在 Phase 2/4 通过 WebSearch 抓取搜索结果与 Google AI Overview（均属于运行时引入的外部网页/搜索内容）。

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches the user-provided target URL (the $ARGUMENTS input) at runtime via mcp__web_reader__webReader or mcp__playwright__browser_navigate and uses the extracted PAGE_TITLE to drive all subsequent queries and prompts, so the external page content directly controls agent behavior (flagged URL: the input URL passed as $ARGUMENTS).