openai-realtime-audio
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill acts as a comprehensive reference for OpenAI's audio-related endpoints. All documented network activity and API interactions are restricted to official and trusted OpenAI domains (api.openai.com, developers.openai.com).
- [SAFE]: Code samples follow security best practices by illustrating the use of short-lived ephemeral client secrets (ephemeral keys) for browser and mobile environments, rather than embedding persistent API keys.
- [SAFE]: The skill relies on standard, reputable software libraries including the official openai SDK for Python and Node.js, alongside common communication libraries like ws and websocket-client.
- [INDIRECT_PROMPT_INJECTION]: The skill documents architectures that ingest untrusted user audio and support high-capability tools such as Function Calling and Model Context Protocol (MCP). Implementation of these features establishes a surface for indirect prompt injection; however, the skill includes guidance on managing model behavior and session parameters to mitigate these risks.
Audit Metadata