recut-worlds

What problem does it solve? Building a reusable creative world (characters, locations, styles, stories) for AI video production requires structured modeling and consistent media generation; without a shared operating procedure, agents produce inconsistent characters, drifting scenes, and uncontrolled canvas edits. ## Core Features & Use Cases - Entity and Relation Modeling: Create, update, archive, and confirm world entities with typed attributes, ordered media properties, and bidirectional relations via the recut.worlds.* tool surface. - Canvas Projection Management: Place entity cards, attribute cards, and arrow edges on the World Canvas so every visible node stays a reference projection of the single source of truth in entity attributes. - Governed Media Generation: Read world references first, attach role-tagged anchors (character, environment, style-ref, voice, storyboard), then submit image, speech, or video generation with videos left pending user confirmation. - Use Case: When asked to generate a scene video featuring the main character, the skill reads the world graph, attaches the character reference image and voice reference with correct roles, submits the video proposal, and places a pending media node on the canvas for the user to confirm. ## Quick Start Ask the agent to read the current world with recut.worlds.get and then create a new character entity with a reference image placed on the canvas.