extraction-pipeline-patterns

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

• Third-party content exposure detected (high risk: 0.75). The required workflow is an extraction pipeline that ingests user-provided files (e.g., PDFs/archives/emails/HTML) and routes their readable body text through format-specific extractors into the agent’s LLM context; if the file content is authored by an outsider, that free text becomes LLM-readable input (indirect prompt injection risk via document text).