detecting-indirect-prompt-injection

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can be ingested at runtime via the --html/--pdf/--image/--text CLI inputs (e.g., downloaded or user-provided third-party artifacts), which the script extracts/OCRs/normalizes into readable text and then passes to the LLM Guard / transformer classifiers (scan_llmguard(normalized) / scan_model(normalized)), creating indirect prompt-injection exposure.