The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructs the user or agent to use administrative privileges for environment setup on Linux systems, specifically calling sudo apt update && sudo apt install -y yt-dlp within the transcript workflow documentation.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content from YouTube which could contain malicious instructions designed to subvert agent behavior.\n
Ingestion points: Video titles, manual subtitles, and auto-generated captions retrieved via yt-dlp in references/transcripts.md.\n
Boundary markers: None present to distinguish between data and instructions.\n
Capability inventory: Shell command execution (yt-dlp, whisper, eyeD3, curl), file system modification (FileUtils.mv, rm), and package management (brew, apt, pip).\n
Sanitization: Employs basic character stripping and replacement (using tr and regex) for generating filenames, which does not mitigate injection into the agent's logic.\n- [COMMAND_EXECUTION]: The skill frequently invokes command-line utilities such as yt-dlp, eyeD3, whisper, and curl using shell execution methods or Ruby's sub-process handling to manage media files and metadata.\n- [EXTERNAL_DOWNLOADS]: The skill fetches external code and resources. It installs the yt-dlp and openai-whisper packages from official registries and utilizes bundler/inline to dynamically install the thor and gum gems at runtime. Additionally, it downloads media content from YouTube and cover art from user-supplied URLs.

youtube