NousResearch/Hermes-Agent — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a self improving AI agent I can run from my terminal or talk to through apps like Telegram, Discord, Slack, WhatsApp, Signal, and email. I want it to feel like a personal assistant that remembers useful things across conversations, can search past chats, creates reusable skills from hard tasks, and gets better as I use it.

It should let me pick different AI model providers without changing code, including common options like OpenAI, OpenRouter, Nous Portal, Hugging Face, or my own endpoint. Include a clean terminal chat experience with streaming output, history, multiline messages, slash commands, and the ability to interrupt and redirect it.

Please add setup commands, a config flow, tool enabling, diagnostics, updates, and a messaging gateway. It should support scheduled automations in plain language, like daily reports or backups, and be able to run locally, on Docker, over SSH, or on a cheap cloud machine. Look up current docs online if you need to.

Want more depth? Deep Reverse