format-specific-extraction

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 when the operating user supplies an outsider-authored document (e.g., DOCX/PPTX/PDF/ZIP/TAR/EML) whose extracted text/metadata is parsed into readable prose at runtime via the described extraction paths (e.g., extractors/pdf/mod.rs → per-page text/OCR, extractors/docx.rs/pptx.rs/odt.rs → XML text, extractors/email.rs → HTML/text body, extractors/archive.rs → plaintext file contents).