The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes common command-line tools for media analysis and workflow automation. It specifically mentions ffmpeg for generating frame grids and yt-dlp for retrieving video metadata. The usage of yt-dlp is constrained to metadata extraction (--skip-download --print), which is a standard, low-risk operation for media workflows.
[EXTERNAL_DOWNLOADS]: The workflow suggests using external data sources and libraries for style mining, including Kaggle CSVs, the schrute R package, and pandas. It also references third-party repositories for lip-sync fallbacks (e.g., MuseTalk), while explicitly advising a safety scan before integrating external code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external content such as show transcripts, Kaggle datasets, and video metadata. While this provides a potential surface for indirect prompt injection, the risk is mitigated by clear instructions to mine for structure rather than copying text, and a mandatory QA gate that validates raw clips and final renders against the original script and product brief.

sitcom-parody-ad-director