@aronhouyu: Loop Engineering - Many people still use Claude Code, Codex, or Cursor like a chatbot: drop a prompt → wait for response → copy it out → fix bugs → drop a new prompt... cycle...

Loop Engineering Now many people, when using Claude Code, Codex, or Cursor, still operate them like chatbots: Drop a prompt → wait for response → copy it out → fix bugs → drop a new prompt... cycle repeats. This repo tells you directly: Stop prompting manually all the time. Instead, design a loop that automatically issues commands to the agent and directs its work. It already provides several ready-made loops, such as: • Daily task routing loop • PR auto-review loop • CI check loop • Dependency update loop • Changelog writing loop • Post-merge auto-cleanup loop • Issue classification loop It also comes with several CLI tools that let you: • Scale the loop • Estimate token usage • Check if your repo is suitable • Add memory/state to the agent • Add manual intervention steps • Add validation gates • Run safely via GitHub Actions In the past, Prompt Engineering was about writing better instructions. Now, Loop Engineering is about building a system that lets the agent run, check, fix, and expand itself, while you just supervise at key points. https://github.com/cobusgreyling/loop-engineering…