The Agent Skills Directory

[SAFE]: The skill provides documentation and code samples for the Patter SDK, facilitating the creation of voice-based AI agents.
[SAFE]: External dependencies and references point to well-known technology services and trusted organizations, including OpenAI, Anthropic, ElevenLabs, Hugging Face, and Cloudflare.
[SAFE]: Secret management instructions advise using environment variables for API keys, aligning with security best practices.
[PROMPT_INJECTION]: The resulting voice agents ingest untrusted audio input from telephone callers, creating a surface for indirect prompt injection. This is an inherent characteristic of voice-interfaced AI agents.
Ingestion points: Incoming audio streams from Twilio or Telnyx carriers.
Boundary markers: None explicitly documented for the real-time voice interaction.
Capability inventory: Built-in tools for transfer_call and end_call, plus any user-defined tools.
Sanitization: Not specifically addressed for the audio processing pipeline.

build-voice-agent