jury

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

• Third-party content exposure detected (low risk: 0.10). The workflow crawls and evaluates user-supplied or discovered URLs via shoot.mjs and audit.mjs, ingesting external website HTML and text content into the LLM-driven critic process.

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

• Potentially malicious external URL detected (medium risk: 0.30). An operational dependency on jsDelivr is established when scripts/shoot.mjs fetches axe-core as a fallback at runtime, representing a common/reputable unofficial transporter.