scientific-schematics

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill includes examples that embed API keys directly (an api_key parameter in sample Python code and a --api-key CLI flag), which would require the LLM to include secret values verbatim in generated commands or code and thus poses a high exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow ingests outsider-provided natural-language prompts directly into the LLM for both Nano Banana 2 image generation and Gemini 3.1 Pro Preview quality review, with no requirement to select any pre-existing specific item first.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill makes runtime API calls to OpenRouter at https://openrouter.ai/api/v1 (used via POST to /chat/completions) and the model's textual review/critique returned by that service is injected into the next-generation prompt (improve_prompt → regenerate), so the external content directly controls agent prompts at runtime.