leandre755/HIVE-MIND — reverse-engineered prompt
Reverse engineered prompt
Build me HIVE MIND, a TypeScript autonomous AI agent framework that can chat with users, remember useful context, plan tasks, call tools safely, and delegate work to temporary helper agents when a job is too big.
I want it to support WhatsApp, Discord, Telegram, and a CLI with the same user identity across channels. It should have a blueprint system so I can define how an agent behaves without hardcoding everything. Add safety checks for inputs and outputs, budget limits, permission checks, memory cleanup, and a smart router that can rotate across multiple AI providers when one is unavailable or rate limited.
Please include support for plugins, browser automation, sandboxed JavaScript tool execution, voice input and output where possible, and storage using Supabase with Redis for events or queues. Make it production ready enough to configure from environment variables, run tests, and deploy. Look up current docs online if you need to.
Want more depth? Deep Reverse