ffmpeg
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill describes a 'Video Analysis (Sub-Agent Pattern)' in Section 15 that creates a vulnerability surface for indirect prompt injection. 1. Ingestion points: Video file frames extracted via
ffmpeg(SKILL.md). 2. Boundary markers: None specified to delineate between media content and instructions. 3. Capability inventory: Extensive shell execution capabilities including file manipulation and network operations viaffmpeg(SKILL.md). 4. Sanitization: No sanitization or validation of visual content described. - [COMMAND_EXECUTION]: Provides numerous shell command patterns for media manipulation, including background process execution and batch processing scripts using
ffmpeg,ffprobe, andGNU parallel. - [EXTERNAL_DOWNLOADS]: Includes examples of downloading HLS streams from remote URLs using
ffmpeg.
Audit Metadata