skills/varghq/skills/media-download/Gen Agent Trust Hub

media-download

Pass

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The scripts/video.sh script is configured to download the yt-dlp executable from its official releases page on GitHub (https://github.com/yt-dlp/yt-dlp) if the tool is not found in the system's PATH. This is a standard convenience feature for standalone tool usage.\n- [REMOTE_CODE_EXECUTION]: The skill downloads a binary and executes it after granting execution permissions with chmod +x. Because the source is the official repository of a reputable and well-known tool, this behavior follows established software distribution patterns.\n- [COMMAND_EXECUTION]: The skill wraps several command-line utilities, including yt-dlp, gallery-dl, ffprobe, and curl, within Bash scripts to perform its primary functions of media downloading and metadata inspection.\n- [INDIRECT_PROMPT_INJECTION]:\n
  • Ingestion points: The skill takes URLs and search queries directly from user input as arguments to its scripts.\n
  • Boundary markers: The shell scripts employ the -- argument separator to prevent user-supplied strings from being misinterpreted as command flags by the underlying tools.\n
  • Capability inventory: The skill possesses the ability to perform network downloads and write files to the local file system (defaulting to the ./downloads directory).\n
  • Sanitization: Search queries used in scripts/pinterest.sh are passed through URL encoding using either jq or Python's urllib.parse to ensure the final request URL remains structurally sound.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 1, 2026, 12:07 PM