voice-to-text-config

Pass

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads Whisper machine learning models from Hugging Face, a well-known service for hosting AI models (specifically the Systran/faster-whisper-* repositories).
  • [COMMAND_EXECUTION]: The skill uses pip install to manage the faster-whisper dependency and executes multi-line Python scripts via python -c to verify model caching and perform transcription tests.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by reading and transcribing voice files from the Telegram inbox.
  • Ingestion points: Reads .oga files from ~/.claude/channels/telegram/inbox in Step 3 of SKILL.md.
  • Boundary markers: The transcription result is printed directly to the output without delimiters or warnings to ignore embedded instructions.
  • Capability inventory: The agent has the ability to execute shell commands and install packages, which could be abused if the transcribed text is interpreted as commands.
  • Sanitization: There is no sanitization or filtering applied to the transcribed text before it is returned to the agent context.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 16, 2026, 09:36 PM
Security Audit — agent-trust-hub — voice-to-text-config