higgsfield-gpt-image-2

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

• Third-party content exposure detected (medium risk: 0.65). SKILL.md itself is static reference text, but its required runtime workflow can fetch and extract live product-page content (outsider-authored web text) as part of the static-ads workflow pattern (“fetch product page → extract benefits/claims/certifications/brand-voice substance”), which would be ingested into the LLM context for copy generation.

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

• Potentially malicious external URL detected (high risk: 0.90). The static-ads workflow explicitly states it fetches the live product page from the product_url entries in the brand product catalog at runtime ("fetch product page → extract benefits / claims ... → use as primary source of truth for copy variations"), so external product_url pages are fetched and their content directly controls prompt generation.