audio-edit
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute ffmpeg commands for audio normalization, denoising, and mixing. These are standard media processing tasks and represent the intended functionality of the skill.
- [INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface by ingesting external media files (e.g., in.mp4, bg.mp3) for processing.
- Ingestion points: User-provided video and audio files.
- Boundary markers: None present for the media content.
- Capability inventory: Subprocess execution of ffmpeg.
- Sanitization: None; however, because the data is processed as binary media by a specialized tool rather than as natural language by the LLM, the risk of instruction injection is negligible.
Audit Metadata