learn-artist

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

• Third-party content exposure detected (high risk: 0.80). The skill autonomously searches and downloads public YouTube content via discover-music (see download_and_separate calling youtube-search and youtube-stems), optionally queries NZBGeek (nzbgeek_fallback), and parses YouTube .info.json / taxonomy-skill outputs to enrich metadata and auto-queue/train models—clearly ingesting untrusted, user-generated third-party content that can change what the agent downloads, queues, and how it classifies/models.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill invokes docker run with the external image "cherrymint/rvc_webui:rvc_boss" at runtime (docker run ... "${DOCKER_IMAGE}"), which will pull and execute remote code from the registry, so this is a required runtime dependency that executes external code.