browser-trace

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

• Third-party content exposure detected (high risk: 0.90). This skill explicitly attaches a read-only CDP client and samplers to arbitrary browser targets (see start-capture.mjs spawning "browse cdp " and snapshot-loop.mjs calling "browse --ws get html body" / taking screenshots) and the SKILL.md and examples show the agent reads DOM dumps, network requests, console exceptions and page content from potentially public/untrusted sites as part of its core debugging workflow, so third-party page content can materially influence analysis and subsequent actions.