hypogenic

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

• Third-party content exposure detected (medium risk: 0.30). Hypogenic ingests outsider-authored free text via the user-provided dataset JSON/text_features and optional PDF-derived literature content (placed/processed from user-supplied files), which then get injected into prompt templates at runtime for LLM generation/inference.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs cloning and running scripts from the external repository https://github.com/ChicagoHAI/hypothesis-generation (e.g., "bash ./modules/setup_grobid.sh" and running pdf_preprocess.py), which fetches remote code and executes it during literature-processing runtime.