page-prep

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

• Third-party content exposure detected (medium risk: 0.65). Yes: scripts/overlay-db.js fetches outsider-authored free text at runtime from public URLs (GitHub raw Rules.json and EasyList hide rules), embeds it into the bundled script (buildBundle), and that bundle is then evaluated in the page via playwright-cli eval, so the LLM context can be influenced indirectly by untrusted fetched content.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's runtime script scripts/overlay-db.js fetches external pattern databases from https://raw.githubusercontent.com/cavi-au/Consent-O-Matic/master/Rules.json and https://raw.githubusercontent.com/easylist/easylist/master/easylist_cookie/easylist_cookie_general_hide.txt and embeds their content into the injectable bundle which directly influences dismiss/hide recipes and therefore agent actions.