graph TD A[Start: Image Prompt Generator] --> B[SILO 1: Style Analysis] B --> C[Task 1: Analyze Reference Images] C -- style_analysis_text --> D[Task 2: Create Style Guide] D -- style_guide --> E[SILO 2: Prompt Generation] E --> F[Task 3: Generate Base Prompts] F -- base_prompts --> G[Task 4: Optimize Prompts] G -- optimized_prompts --> H[SILO 3: Compilation] H --> I[Task 5: Compile Final Collection] I --> J[End: Final Prompt Collection] %% Input connections K[Reference Images] --> C L[Pun Text from Subagent 1] --> F %% Data flow annotations C -- Uses GPT Vision Analysis --> D D -- Structured Style Guidelines --> F F -- Initial Prompts --> G G -- DALLE-Optimized Prompts --> I %% Style definitions classDef silo fill:#e6f3ff,stroke:#333,stroke-width:2px classDef task fill:#f9f9f9,stroke:#666 classDef input fill:#f0fff0,stroke:#333 %% Apply styles class B,E,H silo class C,D,F,G,I task class K,L input