The Agent Skills Directory

[Official SDK Integration]: The skill provides instructions for installing and using the official google-genai (Python) and @google/genai (JavaScript) SDKs. These are established libraries provided by the vendor for interacting with the Gemini API.
[Real-time Data Ingestion Surface]: The skill is designed to handle continuous streams of audio, video, and text data via WebSockets using the send_realtime_input and send_client_content methods. This creates a surface where external data is processed by the agent.
Ingestion points: Real-time user audio, video frames, and text input are received during a live session (SKILL.md).
Boundary markers: The skill demonstrates the use of system_instruction to define the agent's role (e.g., "You are a helpful assistant") which helps establish operational boundaries.
Capability inventory: The skill facilitates real-time dialogue, transcription, and translation, including the ability to call asynchronous tools (NON_BLOCKING function calls).
Sanitization: Standard API guardrails are expected at the protocol level, and the skill encourages best practices like using ephemeral tokens for secure client-side authentication.
[Reference to External Documentation]: The skill includes links to official documentation hosted on ai.google.dev and examples from the google-gemini cookbook on GitHub. It also references well-known technology partners like Firebase, LiveKit, and Software Mansion for WebRTC and streaming integrations. These references are informative and point to established, reputable service providers.

gemini-live-api-dev