browser-intent

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md describes browser_act delegating to page-agent that “turns the DOM into text” and runs an LLM tool-call loop against it, meaning runtime browser-extracted page content (outsider-authored web text) can be ingested into the agent’s LLM context.