The Agent Skills Directory

[SAFE]: The skill documents best practices for managing sensitive credentials (such as OpenAI, Google, and Gemini API keys) by ensuring they are handled server-side and never exposed to the client or browser logs.
[SAFE]: It provides a reference to official Google Cloud documentation for streaming speech-to-text implementation, which is a well-known and trusted service.
[SAFE]: The instructions mandate that all tool calls originating from realtime sessions must pass through a guarded agent executor rather than being executed directly, ensuring security controls like validation, auditing, and redaction are maintained.

voice-transcription