brand-mention-monitor

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

• Third-party content exposure detected (medium risk: 0.30). Brand Mention Monitor’s runtime workflow explicitly “search[es]” and then “extract[s]” third‑party posts/pages on social/review/news platforms (e.g., Step 1 mention sweep and Step 2 deep extraction), so outsider-authored free text from those sources is ingested for scoring and rendered in the triage console.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's HTML template and example explicitly include external CDN URLs that will be fetched at runtime and execute remote JavaScript (e.g. https://cdnjs.cloudflare.com/ajax/libs/d3/7.8.5/d3.min.js and the world-atlas JSON URLs), which the skill mandates keeping and which the rendered triage widget relies on to run client-side code.