meta-criteria-generator

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow takes outsider-authored free text from the user (“title/keywords” and the LLM-generated inclusion/exclusion text derived from it) and passes it into the required extraction step via scripts/extract_criteria.py --inclusion/--exclusion, where the script reads the text and extracts the contents inside {...}.