shobcoder/shob — reverse-engineered prompt

Reverse engineered prompt

Build me a beta version of Shob, an AI agent workspace for developers where I can manage several AI coding sessions in one place instead of bouncing between chats, terminals, and tools.

I want to be able to run multiple agents on the same project at once for things like fixing bugs, writing tests, updating docs, refactoring, and building features. Make it feel safe and controlled, with a clear approval flow before any command runs, files are touched, or code changes are applied. It should understand the current repository, branch, and local changes so the agents work with Git in a sensible way.

I also want subagents, so a bigger task can be split into smaller jobs and rolled back up into one result. Show live progress as tasks are planned and completed, plus terminal output and a code diff view so I can review what happened before accepting changes.

Keep the experience clean and practical, something useful for freelancers, students, small teams, and startups. If you need details, check the current docs online.

Want more depth? Deep Reverse