sprite-from-video

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The scripts video_frames.py and video_loop.py execute the external binaries ffmpeg and ffprobe via subprocess.run to extract and analyze video frames. These calls use list-based arguments without a shell, which is the recommended security practice for preventing command injection. The skill also includes explicit instructions to stop and ask the user for installation if these binaries are missing, rather than attempting to install them automatically.- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied external data in the form of video clips and character stills. While this constitutes an ingestion surface for untrusted data, the processing is handled via deterministic image processing algorithms (chroma unmixing, silhouette detection, and drift measurement) rather than being passed into a prompt for LLM evaluation. The capability inventory is limited to standard media manipulation tools and local filesystem writes to specified output directories.- [EXTERNAL_DOWNLOADS]: The documentation references external client scripts (e.g., comfy_asset_client.py) for the generation phase of the pipeline. However, the skill does not include code to download or execute these from remote sources automatically; it provides the workflow configuration while leaving execution and environment setup as separate manual steps.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 02:45 AM