what is miniature-rescue-prompt fr?

This project is a prompt-writing skill for a very specific style of AI-generated short video: a giant human hand using an everyday household object to rescue a tiny miniature world from disaster, like a flooded toy town or a burning oil field made of models. It is meant to be used with text-to-video AI tools such as Seedance, Kling, Runway, Veo, or Sora, and can be installed as a skill for Hermes, Claude, Codex, Cursor, or similar AI agent tools. Rather than just filling a prompt with cinematic-sounding adjectives, the skill follows a structured process. First it figures out what kind of disaster the user wants, drawn from themes like environmental problems, health metaphors, social and emotional issues, or city life problems. Then it matches that disaster to an everyday tool whose real function logically explains why it would fix the problem, such as a kitchen sponge soaking up a flooded street, or a plate cover smothering an oil well fire by cutting off its air. The skill deliberately avoids weak matches like a fire extinguisher putting out a fire, since those lack the surprising but sensible twist that makes the video interesting. Each clip follows a fixed three-beat, seven-second structure: the miniature people react in panic to the disaster, the giant hand enters and performs one simple action like pressing, dripping, or wiping, and then the disaster clears while the miniature people cheer. The skill recommends a single fixed camera shot per clip, since switching between shots tends to make the AI video model draw inconsistent buildings, characters, or disaster states. The repository includes reference files listing paired disaster-and-tool ideas and visual style rules for keeping the miniature, tilt-shift photography look consistent, avoiding a CG or cartoon appearance. It is aimed at short-video creators and content teams who want to turn abstract topics such as pollution, loneliness, or online harassment into an easy to understand, feel-good miniature rescue clip.