I have compiled some best effort motion canvas skills. They also contain agent instructions for how to capture and export screenshots and the final video, so, in theory, the agent can create a video fully autonomously. Here is a very boring demo: https://www.youtube.com/watch?v=e8eHLTMWLlo