goragodwiriya/claude-code — reverse-engineered prompt

Build me a research friendly terminal coding assistant like the Claude Code snapshot described here. I want to open it in a project folder, chat with Claude about the code, ask it to search files, read files, edit files, write new files, run shell commands, and fetch web pages when needed.

Make it feel safe for a normal developer to use. Before anything risky like editing files or running commands, show a clear permission prompt and let me approve or deny it. Add simple slash commands for things like login, config, doctor checks, cost tracking, review, commit, memory, tasks, theme, and resume.

Use TypeScript with Bun and a terminal UI that feels polished. Include a clean command registry, tool registry, API service layer, session state, and basic plugin or skill support. It doesn’t need to be identical to Anthropic’s app, but it should be runnable, understandable, and useful for studying how an agentic CLI works. Look up current docs online if you need to.