speech-recognition
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements audio transcription functionality, which naturally involves processing untrusted external audio data. This creates an attack surface where audio content could be designed to produce transcripts containing instructions for downstream agent tools.
- Ingestion Points: Audio data is ingested via
LiveTranscriber.appendAudioandtranscribeFile(url:)as described inSKILL.mdandreferences/speechanalyzer-patterns.md. - Boundary Markers: The provided Swift snippets do not demonstrate the use of delimiters or explicit boundary markers when handling the resulting transcript strings.
- Capability Inventory: The skill focus is limited to transcription; it does not include capabilities for executing shell commands, writing to sensitive files, or performing network operations based on the transcript content.
- Sanitization: The implementation patterns do not include text sanitization or validation logic for the generated transcripts.
Audit Metadata