computer-and-browser

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This code implements a native-messaging-backed Chrome extension that intentionally allows a remote MCP server to execute arbitrary JS/CDP commands, read page HTML/text, capture screenshots, and synthesize input — capabilities that enable data exfiltration and act as a powerful remote-control/backdoor for the user's signed-in browser.

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

• Third-party content exposure detected (medium risk: 0.30). extension/commands.js ingests page-originating free text at runtime via commands like page.text (reads document.body?.innerText) and page.pick (captures element.innerText/textContent and outerHTML) from whatever signed-in tab the user/agent has opened or that the agent attaches to.