content-ideator

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow runs Phase 5 “Recreate Brief” on locked ideas, which downloads/transcribes source reels and reads frames/slides (including verbatim on-screen text and captions/transcripts) from externally authored social/video content via /analyze-content and Apify, so outsider-authored free text is ingested at runtime.

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

• Potentially malicious external URL detected (high risk: 0.80). This skill's Phase 5 explicitly downloads a source reel at runtime (the seeded "🔗 url" / Apify "videoUrl" / childPosts[].displayUrl via yt-dlp / Apify actor shu8hvrXbJbY3Eb9W) and injects its transcript/frames into the recreate brief the agent writes, so fetched external content directly controls the agent's prompts/output.