color-correction

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides a shell command template to extract video frames using the cv2 (OpenCV) library. This execution is a functional requirement for the primary purpose of video color analysis.
  • [DYNAMIC_EXECUTION]: The instructions guide the agent to generate and execute a Python snippet for frame extraction. This script generation is simple, transparent, and specific to the media processing task.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests external image and video files, creating a potential surface for indirect prompt injection. However, the primary tool used (get_image) is described as a read-only statistical analyzer that returns numerical data, which limits the effectiveness of instructions embedded in the media.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 01:58 PM
Security Audit — agent-trust-hub — color-correction