The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses execFile to invoke external tools like ffmpeg, ffprobe, yt-dlp, and whisper. This is necessary for video scene detection, keyframe extraction, and transcription. The use of execFile with argument arrays is a secure practice to prevent shell injection by bypassing the shell parser.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading video content from URLs using yt-dlp (Mode C). This is a core feature for analyzing online video content and is properly documented for the user.
[PROMPT_INJECTION]: The skill ingests untrusted external data from subtitles and video transcripts, which constitutes an indirect prompt injection surface. This is a common characteristic of data-processing skills.
Ingestion points: scripts/extract_subtitles.mjs and scripts/transcribe.mjs process external text data from files and audio.
Boundary markers: The analysis workflow uses JSON schemas (references/schema.md) to structure data before presentation to the agent, providing clear boundaries for the processed content.
Capability inventory: Subprocess execution for media processing (ffmpeg, yt-dlp), and file system access for local report generation.
Sanitization: Basic HTML tag stripping is performed on subtitle text to prevent rendering issues or content confusion.

film-breakdown