Loops skill
A hundred ready loops for Claude Code, Cursor and Codex: the agent acts, runs a check and fixes the cause until the goal is met.
- 125 ★ on the repository
- Last commit
Extended access
Why this level
- A loop edits code and runs check commands autonomously, so set the iteration limit explicitly.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/loopsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- A loop edits code and runs check commands autonomously, so set the iteration limit explicitly.
This is third-party code. Review the repository files before installing.
What it does
A collection of 100 loop prompts with Chinese documentation. Each loop sets a goal, a check command, a maximum iteration count and an exit condition, then repeats: check, find the cause, fix the minimum, check again. Scenarios cover development (`test-until-green`, lint and CI repair, TDD), data, content, product work and learning. In Claude Code the call is `/loops <name>`, plural, to avoid colliding with the built-in `/loop`. The `install.sh` script links the repository into Claude Code, Trae, Codex and Windsurf, and writes project rules for Cursor and Cline.
Who it is for. Developers with a checkable goal such as green tests or clean lint who want to start a loop with a ready command rather than writing a prompt each time.
Good fit when
- Tests fail after a refactor and must be driven to green.
- CI is red and the agent reads logs, reproduces and fixes step by step.
- An article or document is polished until a quality score hits a threshold.
- The same scenario is needed in Claude Code, Cursor and Codex.
Not a fit when
- There is no check command: without one the loop cannot tell when to stop.
- The agent must not make many autonomous edits in a shared branch.
- A single built-in `/loop` call for timed polling is enough.
Example request
/loops test-until-green: run npm test, find the cause of the first failure, fix with a minimal diff and repeat, ten iterations at most.Limitations
Documentation is mostly in Chinese, with an English README. A loop helps only with a checkable goal and an iteration cap: without them the agent can keep editing and burn tokens. The plural `/loops` command is needed because of the built-in `/loop`. Some loops target the Trae environment and Chinese services. The repository states no license. The last commit dates from July 2026.
How to disable. Delete the link or folder `~/.claude/skills/loops` and the `~/.loops` directory. If you ran `install.sh`, remove the `# >>> Loops` blocks from Codex `AGENTS.md` and Windsurf rules.
FAQ
How does `/loops` differ from `/loop`?
`/loop` is built into Claude Code and repeats on a timer. `/loops <name>` takes a ready loop from the repository with its own goal and check.
What is a loop made of?
A goal, a check command, a maximum iteration count, an exit condition and execution steps.
What does `install.sh` do?
It links the repository to `~/.loops`, creates links for Claude Code and Trae, and appends instructions for Codex and Windsurf. Re-running duplicates nothing.
Does it work without Claude Code?
Yes, prompts from `prompts/zh/` can be copied into Cursor, Codex, Cline or any other agent.
What if the check never passes?
After the iteration limit the loop stops and asks a human to step in.
Discussion
No questions or reviews yet.
Related
SkillsA Claude Code skill and /looper command that interviews you and builds an agent loop with checks, an independent reviewer and stop rules before it runs
A Codex skill for an autonomous loop: Claude Code computes, Codex reviews hourly, critiques and sends the next prompt.
The author's personal Claude Code toolbox: meta-prompt and todo commands, skills to create skills, hooks and MCP servers, auditors and a Ralph loop
A Claude Code reference plus a collection of over 70 skills for development, security, data and docs, with agent configs
Skills, mods, subagents and hooks for Claude Code: the agent inspects your setup and installs what fits after your approval.
The /codex-workflows skill: Claude Code writes a script, dozens of Codex agents run it, and the run shows on a map
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Loops (this card) | no | 125 | |
| Looper: design agent loops before running them | no | 717 | builds a loop through an interview and an independent reviewer; Loops ships a hundred ready loops for common tasks |
| Open Ralph Wiggum | no | 1,906 | repeats one prompt until a completion marker; here every loop has its own check, limit and exit condition |
README in short
The README explains what a loop is, when to use it, how to install and call it in different agents, and has a catalog of all 100 loops, a scenario lookup table and rules for writing your own.