Awesome Claude Video Skills lists 304 open-source video tools for coding agents

A product team that wants a coding agent to make the next explainer video has had to search GitHub repo by repo and hope. Zhu Yansen's catalogue puts all of that in one list: 304 open-source skills and toolkits for Claude Code, Codex and other coding agents to make or edit video, sorted by the job each one does. The repository holds 676 stars at 73 forks and lists each entry with a security grade.

Writing in the README on GitHub, Zhu groups the 304 entries into ten sections. The counts are the author's own: 51 frameworks and toolkits, 61 promo and demo generators, 49 explainers, 41 editing skills, 37 shorts and social, 20 motion graphics, 10 scripts and learning tools, 10 music videos, 17 stories and animation, and 8 avatar skills. A reader hunting for one particular job starts in that section rather than at a search bar.

The security grade, and what it misses

Each entry carries a security badge from Agent Skills Hub, Zhu says, marked safe or caution. The README does not state the grading method, so treat the badge as a filter rather than a verdict: a safe mark is a reason to try one project before another, not a reason to run it without reading its source. A coding agent with filesystem and shell access is a user that has been granted much of your laptop, and a skill it runs inherits that.

Zhu also names the entry bar for the list: a repo must make or edit video, run as an agent skill, carry documentation, and either meet a star count or clear a quality check. Those four are the thin line between a working directory of 304 options and a dump of every GitHub repo with the word video in its name.

What a reader does with it

Pick the section that matches the job. For an ad agency making a 15-second product shot, that is promo and demo generators or motion graphics. For a learning platform, it is explainers or scripts and learning. For a channel on TikTok, it is shorts and social. The list is a short-list of short-lists.

Reading the catalogue is not the same as adopting any one skill, and that is where the useful question is. Zhu does not state what hosted APIs each skill calls, how much a one-minute render costs, or how long any of them take. A buyer who has to put a video in front of a client picks three candidates from the fitting section, runs each one against a sample scene, and notes the API bill at the end of each test. The catalogue narrows the search; the three test runs decide which one goes into the workflow.

The repository is a working document rather than a finished one, so expect counts to move as entries come in and out. Both the ten sections and the four-item inclusion bar are the stable part to lean on when the number next to each category changes.

Source

Awesome Claude Video Skills, Zhu Yansen on GitHub.