AI Prompt Optimizer and Scene Splitter

Build a clean, minimal base workflow with this logic: 1) Manual Trigger 2) Set node → stores "user_prompt" 3) AI Agent + LLM → optimize prompt into "master_prompt" 4) AI Agent + LLM → split master_prompt into 4 scenes (only scene_1...scene_4, durations fixed at 2.5) 5) Parse JSON safely and store final scenes as variables 6) Final Output: return scene_1...scene_4 DO NOT add Sora API yet. Keep everything very minimal. Only focus on: prompt optimization → scene splitting → valid JSON output. Use clean node names.

English: Build a minimal workflow: Manual Trigger, Set user_prompt, AI Agent optimizes to master_prompt, AI splits into 4 scenes with 2.5s durations, parse JSON and output scenes.