google-gemini-embeddings

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

• Third-party content exposure detected (high risk: 1.00). The skill includes runtime ingestion of arbitrary user/third-party text (e.g., the templates/rag-with-vectorize.ts POST /ingest handler that accepts documents and stores metadata.text for retrieval, plus embeddings-fetch.ts which reads text from query params or POST bodies) and then uses that untrusted content as context for LLM generation, creating a clear avenue for indirect prompt injection.