devbeta-hcode/pig-agents — reverse-engineered prompt
Reverse engineered prompt
Build me a real AI coding workspace, not a fake demo. I want a thin web app with a React front end and a Node TypeScript backend where I can open a local project folder, browse files, edit code, use a terminal, and chat with an agent that can actually read files, search code, make patches, create files, run commands, and browse the web or an embedded browser when needed.
The UI should feel like a lightweight code editor, with a file tree, tabs, Monaco style editor, terminal, chat panel, diff view, and a browser panel. In chat, stream what the agent is doing as it works, including actions, observations, and the final answer. Any shell command should ask for approval per workspace, and web tools should have a separate auto allow setting. After edits, have it automatically run checks like typecheck, lint, test, and build.
It should work with OpenAI or a local OpenAI style server, support project rules from the workspace, and keep all file access sandboxed to the opened folder. Please cite the source and respect copyright, and look up current docs online if you need to.
Want more depth? Deep Reverse