ai-ml

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly lists hugging-face-datasets ("Create and manage datasets on HF Hub") in External Dependencies and includes RAG code that retrieves documents from a vector DB and injects that context into model prompts, which together indicate the agent will ingest public, user-contributed HF Hub datasets (untrusted third-party content) and use that content to influence model outputs and subsequent actions.