video-editing
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions to execute various shell commands using
ffmpegandffprobefor video concatenation, transitions, audio mixing, and caption burning. - [EXTERNAL_DOWNLOADS]: Fetches and installs related agent skills from the PrunaAI repository and the HeyGen Hyperframes project using
npx. - [INDIRECT_PROMPT_INJECTION]: The skill processes external media files and transcription JSON data (e.g., from WhisperX), creating an attack surface where maliciously crafted input data could attempt to influence the agent's behavior during assembly or captioning.
- Ingestion points: Reads transcription data from JSON files and processes user-provided video and audio clips in
SKILL.mdandreferences/captions.md. - Boundary markers: No specific delimiters or safety warnings are provided for the ingestion of transcription content.
- Capability inventory: The skill utilizes shell command execution via
ffmpeg,npx, andnpmscripts across multiple reference files includingreferences/assembly-concat.mdandreferences/combination-hyperframes.md. - Sanitization: There is no evidence of sanitization or filtering of the text content extracted from transcripts before it is used in video overlays.
Audit Metadata