This Agent Skill (Chinese docs) reverse-engineers a reference video into prompts for the MiniMax H3 video model. The agent watches the full clip, restates the action, camera and narrative for you to confirm or correct, then writes text-to-video, per-shot image-to-video or combined prompts. It does not generate video or need an API key. Install by cloning into your Codex or other SKILL.md skills folder.
INSTALL MINIMAX-H3-VIDEO-REVERSE-SKILL
git clone https://github.com/gnipbao/minimax-h3-video-reverse-skill.git \
Copied as written in the repo's README. Check the README for options and requirements.
STACK
MiniMax H3MarkdownPython
“MiniMax H3 视频反推 Skill：真实镜头、首帧与动态提示词、AUTO/T2VA/I2V/DUAL，附中文教程与验证工具。”
About Scripts & learning
Scripts, prompts and learning material about making video with agents rather than rendering tools themselves.