The Agent Skills Directory

[Unverifiable Dependencies & Remote Code Execution] (MEDIUM): The file scripts/download_video.py contains a check_yt_dlp function that automatically installs the yt-dlp package via pip if it is not found on the system. This dynamic installation of unpinned dependencies at runtime is a security risk as it bypasses static analysis and could allow for the execution of malicious code if the package is compromised or substituted.
[Indirect Prompt Injection] (LOW): The skill ingests untrusted data from YouTube (video metadata like titles and descriptions) through the get_video_info function. This data is then printed and processed, creating a surface for indirect prompt injection where an attacker could influence the agent's behavior via video metadata.
Ingestion points: scripts/download_video.py (via yt-dlp --dump-json)
Boundary markers: Absent; metadata is printed directly to the console.
Capability inventory: File system write access and subprocess command execution in scripts/download_video.py.
Sanitization: The script uses list-based subprocess.run calls, which prevents direct shell command injection, but does not sanitize the content of the metadata itself.
[Command Execution] (LOW): The script executes external commands using subprocess.run. While it correctly uses list-based arguments to prevent shell injection, it still relies on external utilities and user-provided URLs to perform its primary function.

youtube-downloader