super-video-maker

Fail

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: HIGHINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external web content by capturing screenshots and recording browser sessions from URLs that may be user-provided or agent-discovered during topic research.
  • Ingestion points: URLs are processed in capture_receipts.py, capture_article.py, and agent_browser_recorder.py using Playwright's navigation functions.
  • Boundary markers: Instructions in SKILL.md define editorial roles and quality gates (e.g., 'Source-deck gate'), but there are no explicit 'ignore embedded instructions' markers used when ingesting web content.
  • Capability inventory: The skill has significant capabilities, including executing shell commands via subprocess for video rendering and performing network requests through various API clients.
  • Sanitization: There is no evidence of URL protocol filtering or content sanitization before the browser session begins.
  • [COMMAND_EXECUTION]: Multiple tools within the skill execute shell commands to leverage system utilities for video editing and quality control.
  • Evidence: Tools such as ffmpeg_qc.py, screen_recorder.py, demo_video_composer.py, and build_reel.py use subprocess.run or subprocess.Popen to invoke ffmpeg, ffprobe, Xvfb, xdotool, and ImageMagick. These calls are a core part of the skill's functionality and appear to be correctly parameterized for their intended use cases.
  • [EXTERNAL_DOWNLOADS]: The skill interacts with third-party service providers to generate and retrieve media assets.
  • Evidence: API clients in heygen_client.py, fal_seedance_video.py, and replicate_video.py download generated video files from official endpoints. These interactions are documented and are necessary for the skill's primary purpose. A request to https://api.groq.com/openai/v1/audio/transcriptions in build_reel.py is used for transcribing audio content, which involves sending data to a well-known service provider.
Recommendations
  • HIGH: Downloads and executes remote code from: unknown (check file) - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Nov 11, 2026, 02:34 PM