audio-analysis
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute ffmpeg and ffprobe for media processing. Commands are constructed using quoted variables, which prevents command injection via crafted filenames.
- [PROMPT_INJECTION]: The skill processes untrusted media files, creating a potential surface for indirect prompt injection. 1. Ingestion points: User-provided video or audio source paths in SKILL.md. 2. Boundary markers: None present in the prompt templates. 3. Capability inventory: File system access via Bash and Read/Write tools. 4. Sanitization: The skill uses specific command flags and filters (grep/awk) to extract numerical data, reducing the likelihood of interpreting file metadata as agent instructions.
- [SAFE]: No network activity, credential usage, or persistence mechanisms were detected. The skill performs localized, read-heavy operations on the source media.
Audit Metadata