Open‑Source Video‑Downloader Skill: One‑Command Downloads with yt‑dlp and AI

z-video-downloader Overview

The z-video-downloader skill extracts video‑download functionality from the earlier z-web-pack suite and reinforces it with yt-dlp to overcome interception failures.

Default Output Layout

Video/Downloads/YYYY-MM-DD-Topic/
├── video-file.mp4
├── video-file.info.json
├── download-report.md
└── download-report.json

Key Features

Automatic input detection distinguishes direct URLs, m3u8 streams, YouTube links, B‑site links, and any platform supported by yt-dlp.

Conservative download strategy prioritises a single video to avoid filling the disk with an entire playlist.

On failure the tool records the error cause and a fallback path in download-report.md/json for later review.

Supports direct download of container formats: mp4, webm, mov, mkv, m4v, flv, ogv.

Handles streaming links such as m3u8 and mpd.

Usage

A single command invokes the skill and performs the full download workflow, producing the directory structure shown above.

Open‑Source Availability

The implementation is fully open‑source; users can clone the repository and run the provided scripts.

Local UI Console

Based on the same scripts, a lightweight UI console can be deployed to a server for collaborative use.

Post‑Download Processing Pipeline

video‑link → local MP4 → transcript → key‑frames → multimodal model analysis → learning webpage

The downstream video‑summarization skill is under active development.