anycap-human-interaction

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly supports URL/web-page review (see "Scenario 1: URL / Web Page Review" and examples like anycap annotate https://staging.example.com) which loads arbitrary public web pages and captures human-recorded narration/annotations derived from those pages that the agent then analyzes and uses to drive code/other actions, exposing it to untrusted third-party content that can influence behavior.