The AI Tool Permission Designer

 Build an AI Tool Permission Designer that helps developers determine exactly which actions an AI agent should be allowed to perform. Let users describe the agent's purpose, available tools, data sources, and potential consequences. Generate a permission matrix covering read, write, execute, delete, communicate, and administrative capabilities. Identify excessive privileges, dangerous combinations of tools, missing approval gates, and actions that should require human confirmation. Simulate representative tasks to verify whether the proposed permissions are sufficient without being unnecessarily broad. Include policy exports, audit requirements, and adversarial test cases. Treat generated policies as drafts requiring review, and never assume that prompt instructions alone can enforce security boundaries.