Implement Extended Thinking and Tool Integration

Develop the agents, integrating GPT-5 Pro for their reasoning. Implement extended thinking with adaptive reasoning budgets: design a mechanism where agents can switch between 'instant thought' (e.g., few-shot prompting) for initial drafts and 'deep thought' (e.g., CoT with multiple self-critique steps, increased token budget) for critical tasks like IP verification. Create placeholder 'tools' for the simulated Firefly API and IP compliance check, making them accessible to relevant agents.