whisper
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- PROMPT_INJECTION (SAFE): The skill does not contain any instructions intended to bypass safety protocols or override agent behavior. The 'initial_prompt' parameter mentioned is a standard Whisper feature used for providing context to the transcription model.\n- DATA_EXFILTRATION (SAFE): There are no hardcoded credentials, API keys, or attempts to access sensitive system files. All network-related links are directed towards legitimate research and development resources (GitHub, Arxiv).\n- EXTERNAL_DOWNLOADS (SAFE): Recommended dependencies (openai-whisper, transformers, torch) are industry-standard packages from trusted repositories. Installation commands provided are for legitimate system utilities like ffmpeg.\n- REMOTE_CODE_EXECUTION (SAFE): No instances of downloading and executing remote scripts (e.g., curl | bash) were found. All code snippets demonstrate local execution of standard Python libraries.\n- COMMAND_EXECUTION (SAFE): CLI examples provided are legitimate uses of the Whisper tool and standard package managers for environment setup.
Audit Metadata