chrome-cdp

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

• Third-party content exposure detected (high risk: 0.72). This skill’s runtime LLM-context exposure is via outsider-authored web page DOM/content: cdp nav <url> navigates to a user-supplied URL and later commands like snap, html, net, click, and eval cause the daemon to read and return accessibility tree / outerHTML / resource timing strings from that loaded page into the tool’s output (which an agent would typically feed into the LLM); no direct outsider text is ingested from files/other humans, but arbitrary page content is.