list-expander

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

• Third-party content exposure detected (medium risk: 0.30). scripts/score-batch.ts ingests outsider-authored free text by taking each company’s live homepage text (from fetchHomepageText/fetchHomepageText-based scraping) and sending it into the OpenAI chat-completions prompt for qualification.

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

• Potentially malicious external URL detected (high risk: 0.90). The scripts fetch live website content and inject that external text into LLM prompts which are sent to https://api.openai.com/v1/chat/completions (and also rely on runtime API responses from https://api.prospeo.io), so untrusted remote content can directly influence model decisions at runtime.