evaluating-llms

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

• Third-party content exposure detected (high risk: 0.90). The skill's RAG evaluation and LLM-as-judge patterns explicitly ingest and evaluate retrieved "contexts" from external sources (e.g., vector DB similarity_search, PubMed abstracts, Google Fact Check, arbitrary benchmark/dataset files and web-retrieved chunks) which the agent is expected to read and judge, exposing it to untrusted third-party content that could contain indirect prompt injections.