founder-product-video

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

• Third-party content exposure detected (medium risk: 0.65). The required runtime workflow parses and analyzes outsider-provided product URL content: it calls capture_website(...) on the user’s product_url (public web page text/metadata extracted into LLM-readable “brief”/analysis via analyze_brief/analyze_media), then those generated results are fed back into the agent’s LLM context for downstream prompting.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches a user-supplied product URL (https://...) at runtime via analyze_brief and capture_website and then uses the fetched page/screenshots to build briefs and seed prompts/overlays, so remote content can directly control model prompts and generation.