groq-inference

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly uses GROQ's compound-beta built-in tools (web_search and visit_website) and URL-based image inputs—see reference/tool-use-patterns.md and templates/groq-client.py (compound_search and analyze_image_url) in SKILL.md—which fetch and ingest arbitrary public web pages and image URLs whose content is then read and used to drive agent tool use and follow-up actions, exposing it to untrusted third-party input.