{"slug": "i-wrote-a-claude-skill-that-sets-up-loop-engineering-in-your-repo", "title": "I Wrote a Claude Skill That Sets Up Loop Engineering in Your Repo", "summary": "A developer named Shivansh Pandey created a Claude Code skill called loop-engineering that automates the process of setting up loop engineering in a repository. Loop engineering replaces manual step-by-step prompting of coding agents with an autonomous system that discovers, executes, verifies, and iterates on tasks until a defined goal is met. The skill scaffolds a state file, maker and checker agents, and triggers like scheduled tasks or GitHub Actions, while requiring user confirmation for any side effects.", "body_md": "For months I used my coding agent the same way most people do: write a prompt, read the reply, write the next prompt. I was the loop. I was the thing deciding what came next and noticing when it drifted off course.\n\nLoop engineering is the fix, and I turned it into a Claude Code skill you can run in one command.\n\n**Repo: github.com/sxivansx/loop-engineering**\n\nLoop engineering is designing the system that prompts the agent for you, instead of prompting it yourself, turn by turn. You define a goal and a stop condition once. The loop then finds the work, does it, verifies it with a separate agent, writes down what happened, and runs again until the goal is met.\n\nIt is built from six parts: a trigger, a state file, a discovery step, a maker, a checker, and a stop condition. That is the whole vocabulary. The skill assembles them for you so you are not wiring it together by hand every time.\n\nInstall it, then run it with the goal of your loop:\n\n```\n/loop-engineering keep the main branch's CI green\n```\n\nFrom there it walks the setup and scaffolds a working loop in your repo:\n\n`LOOP.md`\n\nstate file with three living sections, `Open`\n\n, `Done`\n\n, and `Blocked`\n\n. This is the loop's memory between runs, because the model forgets and the file does not.`.claude/agents/`\n\n: a `loop-maker`\n\nthat does one item at a time, and a `loop-checker`\n\nthat independently verifies the work. The checker is the only one allowed to mark something done.`/loop`\n\nfor on-demand runs, a scheduled task for unattended ones, or a GitHub Action for CI-side loops.What you end up with:\n\n```\nyour-project/\n├── LOOP.md\n├── .claude/\n│   └── agents/\n│       ├── loop-maker.md\n│       └── loop-checker.md\n└── (trigger: /loop, a scheduled task, or a GitHub Action)\n```\n\nNothing with side effects, like a scheduled job, a push, or a deploy, happens without your confirmation.\n\n```\ngit clone https://github.com/sxivansx/loop-engineering ~/.claude/skills/loop-engineering\n```\n\nThat makes `/loop-engineering`\n\navailable in every project. If you would rather commit it to a single repo, clone it into that repo's `.claude/skills/`\n\ninstead. The folder name becomes the command, so keep it `loop-engineering`\n\n.\n\nAn autonomous loop is also an autonomous way to ship mistakes. So the skill refuses to set one up without these:\n\nHere is the line I kept coming back to while building it. Two people can set up the exact same loop and get opposite results. One uses it to move faster on work they understand. The other uses it to avoid understanding the work at all. The loop does not know the difference. You do.\n\nI wanted to stop being the bottleneck for every step without handing over the judgment that makes the code mine. A loop does that when it is built right. It runs the boring cycles, and I stay the engineer who reads what it shipped and owns it.\n\nIf you have been hand-prompting your agent through the same task over and over, this is the thing I wish I had a month ago. Clone it, point it at one annoying recurring job, and keep the guardrails on.\n\n**Get it: github.com/sxivansx/loop-engineering**\n\n*By Shivansh Pandey. The idea of loop engineering was articulated by Peter Steinberger and Boris Cherny, and synthesized by Addy Osmani.*", "url": "https://wpnews.pro/news/i-wrote-a-claude-skill-that-sets-up-loop-engineering-in-your-repo", "canonical_source": "https://dev.to/sxivansx/i-wrote-a-claude-skill-that-sets-up-loop-engineering-in-your-repo-2pok", "published_at": "2026-06-30 08:46:13+00:00", "updated_at": "2026-06-30 08:49:13.461928+00:00", "lang": "en", "topics": ["developer-tools", "large-language-models", "ai-agents", "generative-ai", "ai-products"], "entities": ["Shivansh Pandey", "Claude Code", "GitHub", "Peter Steinberger", "Boris Cherny", "Addy Osmani"], "alternates": {"html": "https://wpnews.pro/news/i-wrote-a-claude-skill-that-sets-up-loop-engineering-in-your-repo", "markdown": "https://wpnews.pro/news/i-wrote-a-claude-skill-that-sets-up-loop-engineering-in-your-repo.md", "text": "https://wpnews.pro/news/i-wrote-a-claude-skill-that-sets-up-loop-engineering-in-your-repo.txt", "jsonld": "https://wpnews.pro/news/i-wrote-a-claude-skill-that-sets-up-loop-engineering-in-your-repo.jsonld"}}