What problem does it solve? When working inside a NomiFun Creative Studio canvas, users need a controlled way to propose structural edits—adding text nodes, moving or resizing nodes, and connecting nodes—without risking accidental deletions, invented asset IDs, or unauthorized media generation. This Skill constrains the AI to produce only validated, reviewable canvas operation proposals from the exact bounded context of the current planning turn. ## Core Features & Use Cases - Bounded-context planning: Reads only the canvasContext in the current planning envelope and treats project, node, connection, and asset identities as exact, never inferring missing nodes. - Canonical operation proposals: Emits a single JSON artifact of kind nomifun.creative-studio.canvas-ops/v1 containing 1–64 operations such as add_node, update_node_data, move_node, resize_node, connect, and disconnect. - Strict safety guardrails: Never deletes nodes, never modifies task config fields, never starts image/video/audio generation, and never claims a proposal was applied. - Use Case: A user asks to add a project-planning text node to their canvas; the Skill responds conversationally and appends one JSON artifact proposing an add_node operation with exact text, format, fontSize, and textAlign fields, which the user must review and apply via "应用到画布". ## Quick Start Ask the assistant to plan a canvas change, such as adding a markdown text node summarizing your project milestones, then review the proposed operations and apply them to the canvas.
