transcribe
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands (
whisperandopen) using an input variable$INPUTderived from user-supplied paths. This creates a surface for command injection if the execution environment does not properly handle special shell characters in filenames. - [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and "clean up" the text generated from audio transcription. Malicious instructions contained within the audio/video file (e.g., spoken commands) will be converted to text and presented to the agent, potentially influencing its behavior during the editing phase.
- Ingestion points: Subtitle text generated by the
whisperCLI and read back into the agent context for cleanup inSKILL.md. - Boundary markers: None. The transcript text is processed directly without delimiters or instructions to ignore embedded commands.
- Capability inventory: Shell command execution (
whisper,open -R), file read/write operations. - Sanitization: No sanitization or validation of the transcribed text is performed before the agent reviews it.
Audit Metadata