Implement GPT-5 Enhanced Scenario Generation

Implement the 'Vulnerability Explorer' Red Team agent. This agent, powered by GPT-5 Pro, should generate a detailed, plausible AI deployment scenario (e.g., an AI managing a specific sector like finance, healthcare, or military) and then, through extended thinking, identify at least three non-obvious catastrophic risk vectors. Ensure the agent dynamically adjusts its reasoning budget based on the complexity of the scenario or the initial feedback on risk assessment. Log the adaptive budget adjustments.