Skip to main content
Back

Open-Source Claude Toolkit Automates Your Entire YouTube Production Workflow

An open-source pack of eleven Claude skills handles scripts, thumbnails, retention analysis and Shorts for a YouTube channel without touching your account.

Repo, Full Access
  • Open-source repo youtube-agent-skill gives Claude eleven skills for running a YouTube channel.
  • Covers scripts, hook scoring, title/thumbnail linting, retention analysis, Shorts cutting, SEO, chapters and channel audits.
  • MIT licensed, free, no signup, no API key, runs on clean Python 3 with no dependencies.
  • Hook scorer is a calibrated heuristic from 74 real hooks, not a view predictor.
  • Deliberately does not publish; every skill ends with a copy block and a ship-or-change prompt.
  • Install via Claude plugin marketplace or by cloning into ~/.claude/skills/.

youtube-agent-skill packages 11 Claude skills for scripting, packaging, editing, analytics, repurposing and channel audits. The MIT-licensed project requires no separate signup, external API key or service connection, though users still need access to Claude. Its GitHub repository has more than 2.4k stars.

The project handles production work up to the upload button while leaving publication to the channel owner. An early X post described the package as an AI agent capable of running almost an entire YouTube workflow for free. The toolkit adds no fee, but Claude access may carry its own cost.

Eleven commands, from hook to audit

Each skill appears as a slash command backed by a SKILL.md instruction file. Several commands also call small Python 3 utilities for repeatable scoring, parsing and validation. Those utilities require no third-party packages.

Command What it does
/yt-script Turns an idea into a script, generates five hooks from 21 formulas and scores them before recording.
/yt-package Checks a title and thumbnail as a pair, flagging truncation, repeated wording and vague language.
/yt-edit Converts a transcript into a timecoded edit decision list covering dead air, filler and retakes.
/yt-retention Reads a retention CSV and identifies early audience loss, sudden drop-offs and steady decline.