video-sdk/web

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

• Third-party content exposure detected (medium risk: 0.30). In-session runtime chat/command/transcription text from other participants is ingested via ZoomVideo event callbacks (e.g., client.on('chat-on-message', ...), client.on('command-channel-message', ...), client.on('caption-message', ...)) after client.join() and is handled as free-form strings.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill contains explicit runtime actions that fetch and run remote code (git clone https://github.com/zoom/videosdk-auth-endpoint-sample.git then bun run start) and an automated "Claude Code" update instruction to fetch external docs (https://developers.zoom.us/docs/video-sdk/web/) which would allow external content to control the agent's update behavior.