speech-to-text

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references installation documentation and related functionality hosted on the inference.sh official GitHub repository.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external audio and video files provided via URLs. This creates a surface where content within the media could be used to inject instructions into the agent's context through the transcription output.
  • Ingestion points: audio_url and video_url parameters used in belt app run commands.
  • Boundary markers: No specific delimiters or safety warnings are implemented to differentiate transcribed text from system instructions.
  • Capability inventory: The skill's execution is limited to the belt CLI tool via the allowed-tools configuration.
  • Sanitization: No data sanitization or validation of the transcription content is performed within the skill definition.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 02:19 PM
Security Audit — agent-trust-hub — speech-to-text