meta-ads-generator

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and scrapes untrusted public content — e.g., the user-supplied product URL, Meta Ad Library/competitor ads, and public review sites (Trustpilot/Google/Amazon/G2/app stores) as required inputs in Step 1–2, then uses verbatim quotes and those findings to build the psychology pillar map and drive prompt/tool behavior for ad generation, so third-party content can materially influence actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill requires and WebFetches the user-provided product URL at runtime (e.g., https://www.goodles.com), scraping copy, reviews, colors and imagery which are then injected into the brand-guidelines and image-generation prompts—so that external URL directly controls the agent's prompts and is a required runtime dependency.