anthropic-rag-runtime

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

• Third-party content exposure detected (high risk: 0.70). Outsider free text can enter the LLM context via the runtime retrieval path: the skill packs d.text from the declared retriever into Anthropic Messages “document” blocks (with citations: {enabled: true}), and that retrieved content may originate from outsider-authored sources (e.g., public web pages or third-party documents) depending on the corpus.