youtube-transcript

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides an installation command for the uv package manager via a remote script (https://astral.sh/uv/install.sh) piped to a shell. This is a standard installation method for this well-known developer tool.
  • [EXTERNAL_DOWNLOADS]: The skill utilizes yt-dlp to fetch metadata and subtitle files from YouTube. It also facilitates the dynamic installation and execution of the openai-whisper package and its models from PyPI/OpenAI for transcription fallback.
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to orchestrate a workflow involving file system operations (mktemp, find, rm), text processing (wc), and external CLI tools (yt-dlp, whisper).
  • [PROMPT_INJECTION]: The skill processes untrusted input in the form of YouTube URLs and video titles.
  • Ingestion points: User-provided YouTube URLs in SKILL.md and video titles retrieved via yt-dlp.
  • Boundary markers: The skill uses a regex check for YouTube domains and a tapestry-validate-url utility to restrict input.
  • Capability inventory: Subprocess execution via Bash tool for downloading and processing; Write tool for saving transcripts.
  • Sanitization: External video titles are processed through tapestry-sanitize-filename before being used in file operations or shell commands.
  • [DATA_EXFILTRATION]: Network operations are performed to fetch content from youtube.com and astral.sh. These operations are consistent with the skill's primary purpose of retrieving video transcripts and managing dependencies.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 04:06 PM
Security Audit — agent-trust-hub — youtube-transcript