marketing-analytics

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime “Research Mode: Analytics Audit Tools” uses agent-browser to open user-supplied live URLs (e.g., https://{domain}, pagespeed.web.dev and search.google.com/test/rich-results?url=...) and then extracts the page “text body,” so outsider-authored content from those pages can be ingested.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs installing and using "agent-browser" at runtime by fetching code from https://github.com/vercel-labs/agent-browser (via an npx install command), which results in executing remote code and is required for the live audit workflow.