peer-review

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime path scripts/generate_schematic.py → scripts/generate_schematic_ai.py ingests the caller-provided free-text diagram prompt (args.prompt) and sends it to an LLM for generation/review without needing the user to select a specific pre-existing item first.

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

• Potentially malicious external URL detected (high risk: 1.00). The scripts call the OpenRouter API at https://openrouter.ai/api/v1 at runtime (requests.post to f"{self.base_url}/chat/completions") and use the model responses (image generation + critique) to iteratively modify prompts, and the skill requires an OPENROUTER_API_KEY, so external responses directly control prompt generation.