What problem does it solve? Building production voice agents involves coordinating speech recognition, synthesis, turn detection, and streaming audio across multiple providers, where latency, interruption handling, and privacy failures are common and hard to diagnose. ## Core Features & Use Cases - Architecture Guidance: Separates transport, audio processing, turn detection, transcription, orchestration, tools, synthesis, and observability into explicit components with defined codecs, buffering, and reconnection behavior. - Conversational Behavior Rules: Covers barge-in, silence, noise, partial transcripts, dropped connections, and confirmation for sensitive or irreversible actions. - Privacy, Safety & Evaluation: Enforces recording consent, data minimization, regional routing transparency, and structured testing across devices, accents, networks, and provider throttling. - Use Case: When building a telephony voice agent that must handle interruptions and escalate to humans, use this Skill to define the latency contract, cancellation logic, and evaluation metrics before writing provider SDK code. ## Quick Start Use the voice-ai-development skill to design the architecture and turn-taking behavior for my real-time customer support voice agent.
