collage-broll

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes ffmpeg and ffprobe for image and video processing, which are standard for media creation workflows. Evidence includes tiling images for contact sheets, scaling frames, and removing audio tracks.
  • [COMMAND_EXECUTION]: The script scripts/run_gate3.py executes a wrapper command aigc-video-gen via subprocess.run(). This is used to schedule video generation jobs based on a JSON manifest. The inputs are sanitized through JSON parsing, and the tool itself is a documented requirement in the frontmatter.
  • [EXTERNAL_DOWNLOADS]: The skill references standard API endpoints and documentation from Volcengine (ByteDance) and SiliconFlow for image/video generation. These are reputable services in the AIGC ecosystem and their use is consistent with the skill's primary purpose.
  • [PROMPT_INJECTION]: The skill explicitly defines boundary markers and safety protocols, such as three distinct 'Gate' approvals (Metaphor, Still Frame, and Video), ensuring that the agent does not proceed with resource-intensive or final actions without explicit user consent ('can', 'pass', 'all pass').
  • [REMOTE_CODE_EXECUTION]: No remote code execution patterns were detected. The scripts included are for environment checking and local process orchestration.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 01:49 PM
Security Audit — agent-trust-hub — collage-broll