whisper-extract

Pass

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the whisper command in the shell using user-controlled parameters, including <file_path>, <model>, and <language_code_or_auto>. While the file path is enclosed in double quotes in the template, if the agent's execution environment does not properly escape these variables, a malicious user could provide a path like audio.mp3" && [command] && " to execute arbitrary code.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via the audio/video content it processes.
  • Ingestion points: Raw text extracted from audio/video files via OpenAI Whisper in Step 2.
  • Boundary markers: None identified. The transcript is passed directly to the summarization prompt without delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill has shell execution capabilities (the whisper command) and file-writing capabilities (Write).
  • Sanitization: No evidence of sanitization or filtering of the transcript text before it is used for summary generation in Step 3.
Audit Metadata
Risk Level
SAFE
Analyzed
May 16, 2026, 05:09 AM
Security Audit — agent-trust-hub — whisper-extract