portfolio-ingest

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

• Third-party content exposure detected (high risk: 0.85). Outsider-authored free text from user-provided documents in the project’s uploads/ folder (e.g., PDF/DOCX/PPTX/XLSX that may contain arbitrary content) is extracted at runtime via document-skills:* and then used for entity/context extraction and confirmation in the agent’s LLM flow, creating an indirect prompt-injection pathway.