locators-scanner

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

• Third-party content exposure detected (high risk: 1.00). This skill's Step 2 "Fetch & Parse" explicitly instructs the agent to call a browser MCP to navigate to arbitrary user-provided URLs and capture document.documentElement.outerHTML (i.e., fetch and parse public webpages), so untrusted third‑party HTML would be read and directly drive locator selection and generated code.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly navigates at runtime to a user-provided target URL (e.g., https://example.com/login) using a browser MCP and captures the rendered HTML (document.documentElement.outerHTML) which is then injected/parsed to drive locator generation, so fetched external content directly controls the agent's behavior.