marketing-pricing

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s Research Mode explicitly uses agent-browser to open and ingest freeform text from outsider-authored web pages and review/comment text (competitor pricing pages, G2/Capterra reviews, Reddit search results, and Wayback-cached pages) via agent-browser get text body after opening attacker-influenced URLs/paths.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's Research Mode requires installing and running the external agent-browser skill via "npx skills add https://github.com/vercel-labs/agent-browser", which fetches and executes remote code at install/runtime and is referenced as a required dependency.