scrape-analyze-page

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

• Third-party content exposure detected (high risk: 0.85). The skill reads and LLM-ingests saved HTML that may contain outsider-authored free text (e.g., arbitrary page content from a URL the operating user didn’t author), via clean_html.py PAGE.html → extract_metadata.py PAGE.html and then “Read only the cleaned HTML and metadata” for extraction.