anthropic-prompt-caching-and-context-runtime

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context via the per-request “retrieved per-request ctx” / “user input / variables” portions of the Messages API prompt (i.e., runtime user-provided or retrieved content), which are explicitly not cache breakpoints but are still ingested as readable text.