What you get
An executable directing plan, reusable motion components, and delivery checks backed by frame-level evidence.
Video Production Skills: From Motion Direction to Frame-Level QA
Added 2026-09-03Four independently installable video skills: direct the motion language from a topic, verify a reference remake, create a dark SaaS spot, or package a black-and-white typing opener for code-driven workflows such as HyperFrames and Remotion.
ai-motion-director is the directing layer: it turns a topic, article, script, or brief into a visual metaphor, beat graph, component plan, and asset plan, while keeping an anti-PPT gate active throughout production. reference-video-replica-qc separates five materially different fidelity claims and requires asset, runtime, and delivery gates backed by full-frame, timing-offset, boundary-frame, and archive evidence; proven motion can then be captured as reusable HyperFrames or Remotion components. The two generative skills cover a dark SaaS product language of black space, purple floor light, floating UI, and kinetic type, plus a black-background white-text opener with character-by-character timing and synchronized key clicks. Those two promise style-level creation, not pixel-exact replication. The repository works with Codex, Claude Code, Cursor, and the Skills CLI, and each top-level skill can be installed on its own.