marketing-carousel-growth-engine

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

• Third-party content exposure detected (high risk: 1.00). The skill's Phase 2 "Website Scraping" (analyze-web.js using Playwright) explicitly navigates target URLs and internal pages to extract brand content, testimonials, competitors, and stats from arbitrary public websites, which the agent then uses to drive generation, publishing, and scheduling—exposing it to untrusted third-party content that could inject instructions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill uses Playwright to fetch a user-provided target website URL at runtime (via analyze-web.js), and the scraped site content (features, testimonials, CSS/colors) is directly injected into the slide-generation prompts that control Gemini image/text generation — flagged URL: "user-supplied target website URL (fetched via Playwright at runtime)".