What is designing-loops?

This project is a skill for Claude Code, the AI coding assistant, that helps it decide how to structure a long-running or repeated task and write stop conditions that actually work. When you ask Claude Code to babysit a pull request, run something overnight, or fix a long list of errors, this skill kicks in behind the scenes to figure out which of four loop types fits the job, then writes a ready-to-run plan that includes clear conditions for when the work is finished. The four loop types it chooses between are turn-based, goal-based, time-based, and proactive, an idea drawn from the Claude Code team's own writing on the subject. The skill adds practical lessons learned from watching real multi-hour autonomous runs and checking afterward whether they actually did what they claimed. Its core advice includes always writing the stop condition first, since a number is more reliable than a vague description, and making sure every loop can stop either on success or on a bounded failure. It also warns against letting a long-running task quietly redefine what counts as done under pressure, and recommends that heavier overnight jobs keep their state in a saved file rather than only in the prompt, using a template the project provides. Installation is a single git clone command into a specific Claude Code skills folder. Once installed, the skill also knows when a task is not actually loop-shaped, such as polishing a README, and will say so rather than forcing an unnecessary repeated process. The project is released under the MIT license, so it can be freely used, copied, and modified.