youtube-chapter-clipper

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • Unverifiable Dependencies (MEDIUM): The skill references a local script scripts/smart_edit.py which is not provided for inspection. This script is intended to handle core logic including chaptering and execution. Furthermore, the skill instructions recommend installing system binaries (yt-dlp, ffmpeg) using brew install, which constitutes system-level changes outside the agent's restricted environment.
  • Indirect Prompt Injection (LOW): The skill possesses a clear attack surface for indirect prompt injection.
  • Ingestion points: The agent is instructed to read and perform 'AI analysis' on the full content of external .vtt subtitle files downloaded from YouTube (Step 5).
  • Boundary markers: There are no instructions to use delimiters or warnings to ignore embedded instructions within the subtitle text.
  • Capability inventory: The agent has the capability to execute shell commands (ffmpeg, yt-dlp, python) and write to the filesystem.
  • Sanitization: While basic filename sanitization is mentioned for outputs, no sanitization or safety filtering is applied to the subtitle content before it is processed by the LLM.
  • Command Execution (LOW): The skill relies heavily on executing shell commands with user-provided or externally sourced inputs (YouTube URLs, file paths). While it suggests some safety measures (replacing spaces with underscores), complex command interpolation still carries inherent risks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:15 PM