paper-illustration

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

• Potentially malicious external URL detected (high risk: 1.00). The skill makes direct runtime POST requests (curl) to the Google Generative Language endpoints (https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-preview:generateContent?key=$API_KEY and https://generativelanguage.googleapis.com/v1beta/models/gemini-3-pro-image-preview:generateContent?key=$API_KEY) whose returned text/specifications are saved and injected back into the workflow to control prompts and rendering, so the external content directly controls agent behavior.