5dive-com/5dive — reverse-engineered prompt
Reverse engineered prompt
Build me a Linux command line tool called 5dive that lets me run a small team of AI agents on my own server and talk to them like workers.
I want to install it with one curl command, then create agents like coder, writer, or pm using Claude, Codex, Grok, Gemini, or similar official AI CLIs. Each agent should run in the background as its own Linux user with systemd, keep working after I close SSH, and be easy to start, stop, restart, remove, and inspect with logs.
Agents should be able to send messages to each other through the same CLI, and I should be able to ask an agent a question and get a reply. Add shared account login profiles, basic auth status commands, safe isolation levels, a doctor repair command, a live watch view, and optional Telegram pairing so I can message an agent from my phone.
Keep it simple and private, no external broker, no telemetry, just local files and the server I control.
Want more depth? Deep Reverse