# Loops! (Agent Loop Gallery) loops! (loops.elorm.xyz) describes itself as a gallery of "pre-built agent loops". Its tagline is "Stop prompting. Start looping." The site offers "closed-loop workflows for coding agents", where "each loop includes triggers, feedback gates, and exit conditions so agents self-pace until the job is done." Put simply, it's a recipe book for [[Loop Engineering]]. You don't get a prompt for one step; you get a small program, written in plain language, that tells the agent what "done" means, how to check it, and when to stop. ## What a loop looks like Every loop is a structured kickoff prompt. Here's the shape of the most popular one, "Ship PR Until Green": ```text Goal: PR is open with all CI checks passing Max iterations: 10 Between iterations run: gh pr checks Exit when: all PR checks are success Step 1: Implement the change, test locally, push, open PR, and fix CI until green. Self-pace this loop. After each iteration, run the check command, read the output, and only continue if the exit condition is not met... ``` So: a goal, a hard iteration cap, a check command run between passes, an exit condition the command can prove, and the steps. Loop pages also add anti-cheating rules: don't modify the check command or the exit criteria to force success, don't skip or bypass checks, and if you're stuck after several iterations, stop and report blockers "instead of gaming metrics". Good call: an agent told to "make the tests pass" can always edit the tests instead of the code. Each loop has a trigger type: - **Manual**: you paste the kickoff and the agent runs until the exit condition passes (Ship PR Until Green, Independent Verifier Pass, Ralph Story Executor...) - **Event**: hook-based loops that fire on agent events, e.g. Pre-Commit Guard runs the tests before any `git commit` and blocks it when the suite is red - **Interval**: meant for Cursor's `/loop 5m <prompt>` style ticks, e.g. CI Failure Watcher or Deploy Verification Loop The sitemap listed 40 loops when I checked (October 3, 2026), grouped into categories such as CI, Testing, DevOps and Automation: fix CI until green, coverage until threshold, flaky test triage, dependency upgrades one by one, OpenAPI sync until valid, a "de-sloppify" pass, a [[Ralph Loop]] story executor that works through `.ralph/prd.json` one story per iteration... ## How it works The site is careful to say what it doesn't do: it never runs code on your machine. You get two separate things: 1. **The kickoff prompt** (always). **Use loop** copies it; **Open in Cursor** and **Open in Claude Code** (via `claude-cli://`) pre-fill it through deeplinks. The kickoff is self-contained and works in any agent 2. **Install files** (sometimes). A zip with `.cursor/` and `.claude/` folders (e.g. `.cursor/hooks.json`, `.claude/settings.json`). You extract it at your repo root and restart the agent so the [[Claude Code Hooks|hooks]] load Their split: hooks decide *when* the agent wakes up; the kickoff decides *what* it does on each pass. The site says it "works best with Cursor and Claude Code" and lists [[OpenAI Codex|Codex]], [[Gemini CLI]] and [[OpenCode]] as compatible. You can browse and copy without an account; signing in with GitHub lets you save favorites and submit your own loops (community submissions "may be reviewed before publication"). Loop pages show view and copy counts (around 2K copies for Ship PR Until Green). ## Who made it It's built by **Prince Elorm** (elorm.xyz, `akarikev` on GitHub, `@elorm_elom` on X), a software engineer from Ghana who builds AI solutions for African languages at All Labs and ships lots of side projects. loops! has a sibling site, prompts! (prompts.elorm.xyz), a directory of copy-paste prompts for coding agents. ## Status and maturity - A one-person side project (the footer says "© 2026 loops! Built by elorm") - Free to use, funded by a "Buy me a coffee" link and partner ad slots. A banner says "Loops is going viral and we need your help to keep the project free and alive" (their claim; I found no Hacker News or other coverage to confirm it) - No public source repository that I could find, no changelog, no launch date - The terms say loops are provided "as-is" and "without warranty", and that you're responsible for reviewing agent actions in your own repositories ## My take The value here is the format, not the catalog. Goal, check command, exit condition, max iterations, "don't game the metric": that's a reusable template for ANY loop you write yourself, and a practical take on the trigger, goal and stop conditions described in [[Loop Engineering]]. Two caveats though. The commands are very npm/GitHub-centric (`npm test`, `gh pr checks`), so expect to adapt most loops. And the install bundles are hook configs that run shell commands on agent events, so read them before you extract anything into a repo. Treat them like any script from the internet. ## References - Home page: https://loops.elorm.xyz/ - About ("How loops work"): https://loops.elorm.xyz/about - How to install and run loops: https://loops.elorm.xyz/install - Example loop page, Ralph Story Executor: https://loops.elorm.xyz/loops/ralph-story-executor - Terms: https://loops.elorm.xyz/terms - Sitemap (loop count): https://loops.elorm.xyz/sitemap.xml - Author's site: https://elorm.xyz/ - Author's GitHub profile: https://github.com/akarikev - Sibling site, prompts!: https://prompts.elorm.xyz/ - Hacker News (Algolia API search for "elorm.xyz"): no submissions or comments found on 2026-10-03 ## Related - [[Loop Engineering]] - [[Agentic loops]] - [[Ralph Loop]] - [[Ralph Wiggum Technique]] - [[Claude Code Hooks]] - [[Claude Code]] - [[Cursor]] - [[Continuous Integration (CI)]] - [[Prompt Engineering]]