procurer

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow (Stage 4 “Search with grounding discipline”) fetches public web pages for provider price citations, and the fetched page text can be ingested into the agent’s LLM context, creating an indirect prompt-injection surface from outsider-authored web content.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill explicitly requires runtime web fetches to ground prices and inject supplier pages into the agent context (e.g., https://www.homecenter.com.co/, https://www.sodimac.com.co/, https://www.tecnoglass.com/ and marketplace pages like MercadoLibre referenced in the examples), so those remote URLs will be fetched during runs and their content directly influences the agent's recommendations (prompt/context injection risk).