youtube-transcript

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill includes the yt-dlp -U command, which initiates a self-update mechanism that downloads and executes code from the tool's official update channel.\n- [DYNAMIC_EXECUTION]: The skill uses a shell HEREDOC to dynamically generate and execute a Python script to process JSON transcript data into plain text.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from YouTube video URLs and metadata, creating a surface for potential injection.\n
  • Ingestion points: The skill accepts a user-provided video URL and fetches external metadata including channel names and video titles (SKILL.md).\n
  • Boundary markers: External metadata is interpolated into filenames; the skill does not use specific delimiters to isolate this untrusted content from the surrounding prompt logic.\n
  • Capability inventory: The skill utilizes shell execution (curl, yt-dlp, python3) and file system writes (SKILL.md).\n
  • Sanitization: Filenames are sanitized using the tr command to strip non-alphanumeric characters, mitigating some injection risks.\n- [COMMAND_EXECUTION]: The skill relies on CLI tools for its core logic, including curl for network requests to the DeepAPI endpoint and jq for parsing JSON responses.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 03:46 AM
Security Audit — agent-trust-hub — youtube-transcript