mobile-debugging

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs loading and inspecting arbitrary public webpages and third-party scripts (e.g., Eruda/vConsole bookmarklets loading from https://cdn.jsdelivr.net and https://unpkg.com, and Playwright/Puppeteer examples using page.goto(url) to capture console logs, plus cloud testing flows where you enter arbitrary URLs), meaning the agent would fetch and interpret untrusted web content as part of its workflow.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill includes bookmarklets and CDN script tags that fetch and execute remote JavaScript at runtime (e.g., https://cdn.jsdelivr.net/npm/eruda and https://unpkg.com/vconsole@latest/dist/vconsole.min.js), which directly executes external code in the target page.