youtube-digest

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads YouTube video media and metadata using the yt-dlp tool as part of its media digestion pipeline.
  • [EXTERNAL_DOWNLOADS]: The skill fetches external attachments and slide decks from arbitrary URLs discovered in video metadata using a streaming fetch implementation with byte-size limits to prevent resource exhaustion.
  • [EXTERNAL_DOWNLOADS]: The skill performs shallow clones of repositories from GitHub to analyze project structure and frameworks, utilizing proper argument termination and trusted sources.
  • [COMMAND_EXECUTION]: The skill executes media processing tools like ffmpeg and ImageMagick via subprocesses for scene detection, frame extraction, and contact sheet generation.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes untrusted data from YouTube transcripts and comments. Evidence: Ingestion points (run-transcript.js, run-harvest.js); Boundary markers (hypotheses framing present in SKILL.md, but no technical delimiters); Capability inventory (spawn and spawnSync calls across extraction scripts for CLI tools); Sanitization (path sanitization and markdown escaping implemented).
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 07:56 PM
Security Audit — agent-trust-hub — youtube-digest