comfyui-video-production

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes multiple scripts and logic paths designed to execute system-level commands for workflow orchestration. references/instance-management.md uses subprocess.Popen to start ComfyUI instances and the psutil library to manage and terminate processes. references/concatenation.md utilizes subprocess.run to invoke ffmpeg and ffprobe for video processing. Additionally, SKILL.md provides Bash and PowerShell scripts to automate the launching of multiple local service instances.
  • [DYNAMIC_EXECUTION]: The skill dynamically constructs executable commands at runtime based on user configuration. In references/concatenation.md, the generate_crossfade_command function assembles an FFmpeg command string by concatenating lists of video file paths and calculating temporal offsets. The VideoConcatenator class also builds command lists for subprocess.run dynamically, which could pose risks if input filenames are not strictly validated.
  • [INDIRECT_PROMPT_INJECTION]: An attack surface exists in the evaluation tooling provided with the skill. Ingestion points: The eval/run-eval.sh script reads test case prompts from text files. Boundary markers: None are present; the prompt content is interpolated directly into the command-line call. Capability inventory: The script executes the claude CLI with the interpolated prompt, allowing the data to influence the AI's execution. Sanitization: The script does not appear to validate or sanitize the prompt content before interpolation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 01:03 AM
Security Audit — agent-trust-hub — comfyui-video-production