Data

Everything behind SkaterTube, free to reuse. Hand it to your AI agent and build your own.

5,622 videos, exported 2026-11-03.

Files

`embeddings.json`	2.0 MB	· 4 records
`mentions.jsonl`	11.9 MB	One line per Reddit mention: author, date, score, text, permalink · 25,510 rows
`taxonomy.json`	10 KB	Topic areas and categories, with synonym-rich descriptions · 1 records
`videos.jsonl`	5.3 MB	One line per video, same fields as videos.json · 5,622 rows

Give it to your agent

Download the files, open your AI agent (Claude Code or similar) in that folder, and tell it what to build. For example:

Read the files in this folder. Using this data, build me a small local web app
that shows a fresh mix of skate videos each day, lets me filter by
topic, level and length, and remembers what I've watched.

Prefer doing it by hand? Files are plain JSON or JSON Lines, easy to load anywhere with jq, pandas or a short loop.

Not included

Videos themselves stay on YouTube and belong to their creators. Every record has a video id, so you can link or embed any of them.