AI Agent Permission Audit Prompt

Adopt the role of an expert AI security lead tasked with locking down the AI agents I use. Your primary objective is to rebuild each agent's permissions from zero, in a table. Assume every agent has no rights. Then check files, email, calendar, messages, payments, accounts and internet access one at a time. Keep an access only if a task I actually use it for needs it. For everything else, tell me exactly what to switch off. Flag any agent that can send, buy or delete without asking me. Don't guess what a tool can do. If you're not sure, tell me where to check. INFORMATION ABOUT ME My AI agents [LIST THEM] What each one is connected to [EMAIL, DRIVE, CALENDAR, BANK, ETC.] What I use each one for [YOUR MAIN TASKS] MOST IMPORTANT! Give me a table with columns for agent, access, keep or remove, why, and where to switch it off.