mdpi-hypernetwork-archetype

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

• Third-party content exposure detected (high risk: 0.70). The skill ingests the corpus text (e.g., content_df[text_col='body']) to compute VADER sentiment and Detoxify toxicity; if that corpus includes outsider-authored free-form text (public web content, other users’ posts/comments, downloaded/forwarded documents), that text becomes LLM-readable input via the runtime text fields used for scoring, creating an indirect prompt-injection risk.