cartesia-line-voice-agent

This module implements real-time microphone (MediaStream) capture, converts audio to 16-bit PCM, base64-encodes it, and streams it continuously over a WebSocket as JSON messages. No direct malware indicators are present in the fragment (no dynamic execution, persistence, or hidden behavior), but the capability to exfiltrate sensitive voice/audio content over the network is strong. Actual security posture depends on consent/enablement controls, WebSocket destination validation/allowlisting, transport security (TLS), and how authentication tokens are managed in the surrounding code.