Gitlawb/zero — reverse-engineered prompt

Reverse engineered prompt

Build me a terminal based AI coding assistant I can actually own and run on my machine. I want to launch it with a simple zero command, chat in a clean terminal UI, point it at a repo, let it inspect files, edit code, run shell commands, and help me plan and make changes.

Let me choose which model and provider to use, including both cloud APIs and local models, and make the safety controls really clear so file writes, shell access, network access, and anything outside the project ask for permission unless I explicitly allow more. I also want a non interactive mode for scripts and CI with useful exit codes and JSON output, plus durable local sessions I can search, resume, or fork later.

Please make first run setup easy, include commands to configure providers, list models, and run a doctor style check, and keep context and session data stored locally with no telemetry. If it makes sense, leave room for extra tools or integrations later. Look up current docs online if you need to.

Want more depth? Deep Reverse