loop-engineering

Why we included this project

Once you've moved past hand-feeding a coding agent one prompt at a time, the next step is designing a loop the agent can run on its own. That's what this project is for: a collection of starter patterns and small command-line tools that work across Grok, Claude Code, Codex, and other agent setups. It ships a daily-triage loop you can scaffold, an audit command that scores how well your agent loops are configured, a cost estimator for keeping token spend visible, and helpers for state, memory, and isolated git worktrees. Teams building repeatable, observable agent workflows get concrete scaffolding and checklists here instead of abstract advice, whether they're architects designing agent systems or developers who run coding agents every day.