expert-video

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of subprocess.run across multiple scripts (assemble.py, clip-trim.py, audio-mix.py, deck-render.py, etc.) to execute system binaries like ffmpeg, ffprobe, and node. These executions are essential for its primary function of media processing and rendering.
  • [DYNAMIC_EXECUTION]: Several instances of dynamic code or process manipulation were identified:
  • motion-graphics.py uses os.execv to redirect subcommands to the newer visual-render.py script. This is used for process replacement within the local skill environment.
  • avatar.py and liveportrait.py use importlib.util to dynamically load modules from relative file paths within the skill directory, primarily for maintaining backward compatibility with legacy versions.
  • run-stage.py is designed to execute whitelisted local wrappers (video-producer, avatar, deck-render) to provide progress tracking and recovery mechanisms for long-running production stages.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface as it consumes untrusted data from brief.md and voiceover.md. This data is used to generate prompts for external AIGC video and image tools. The risk is mitigated by the inclusion of mandatory validation gates (GATE A and GATE B) which require manual oversight of the generated scripts and assets before final production.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 06:04 PM