webvoyager

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

• Third-party content exposure detected (high risk: 0.95). The skill’s runtime workflow can ingest outsider-authored free text when it navigates to a runtime-supplied task.startUrl/action.value and then reads await this.page.content() (full HTML text) and page.screenshot() for vision-based prompting, i.e., public web page content authored by others.