speech
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides documentation for integrating speech recognition capabilities using developer-facing SDKs. All mentioned services and tools (DeviceSession, MockDeviceKit) align with the expected development environment for wearable integrations.
- [DATA_EXPOSURE]: The skill requests microphone access which is consistent with its stated purpose of speech recognition. It includes best-practice instructions for handling permissions, such as checking status silently before prompting the user.
- [COMMAND_EXECUTION]: The code samples provided are standard Swift API calls for session management and event listening. There is no evidence of arbitrary command execution, shell access, or unsanitized input processing.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing external data (voice transcriptions). However, the implementation shown is limited to UI display (showPartialTranscript) and local processing (commitTranscript), with no autonomous tool execution or exploitable capability paths identified in the static instructions.
Audit Metadata