audio-to-video

Pass

Audited by Gen Agent Trust Hub on Apr 9, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The scripts/analyze_audio.ts script incorporates user-provided text from the specificRequest variable directly into the system prompt for the Gemini AI model. This creates a surface for indirect prompt injection, as malicious input could attempt to override the storyboard generation instructions. Furthermore, the skill establishes a multi-step chain where the output from Gemini (the generated storyboard and prompts) is used to drive the video generation in the next phase.
  • [EXTERNAL_DOWNLOADS]: The scripts/generate.ts script downloads the final generated video files from fal.ai storage. This is part of the core functionality and interacts with a well-known AI infrastructure provider.
  • [COMMAND_EXECUTION]: The skill includes scripts/split_collage.sh, a bash script that uses ImageMagick to perform image manipulation. While the operations are limited to cropping, the script executes on user-specified paths. The skill also utilizes npx ts-node to run its core logic.
  • [DATA_EXFILTRATION]: The scripts hardcode absolute file paths (e.g., /Users/aviz/.claude/skills/...) to locate the .env file containing API keys. This practice reveals information about the author's local directory structure and system username.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 9, 2026, 12:02 PM