Guide · Agent skill

How to use the Recurring Loop Scheduler agent skill (loop)

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

The loop skill lets an agent run a prompt right away and then check back on it, either on a fixed schedule such as every 5 minutes or at intervals the agent chooses itself. It suits watching a build, a deploy or a pull request without checking by hand. This guide explains what the skill does, how to install it, how to trigger it, and what to watch for so recurring jobs do not keep running unnoticed.

Recurring Loop SchedulerQwenLM/qwen-code · Apache-2.0 Download the skill

Quick start

1Download and unzip. Download the ZIP from this skill's page and unzip it into ~/.claude/skills/ or your project's .claude/skills/ folder.
2Confirm the skill is available. Make sure your agent runtime provides the cron_create, cron_list, cron_delete and loop_wakeup tools before you start a loop.
3Start a loop. Type /loop followed by a prompt, with an optional interval such as 5m or every 30m.
4Check the confirmation. Read the cron expression, cadence and job ID the agent reports so you know what was scheduled.
5Review active jobs. Run /loop list to see which jobs are running.
6Cancel when finished. Run /loop clear to cancel all jobs, or delete a single job by its ID.

What is an agent skill and what does loop do?

An agent skill is a packaged set of instructions and supporting files that an AI agent can load and follow for a specific kind of task. The loop skill is one of these, bundled with qwen-code under Apache-2.0.

When you type /loop followed by a prompt, the agent runs that prompt immediately. It then schedules follow-up checks. With an interval, it creates recurring scheduled jobs. Without one, it decides when to check again, up to about an hour between checks, and stops when the task is complete or blocked.

With an interval (for example /loop 5m or /loop check the PR every 30m): recurring scheduled jobs that keep firing until cancelled.
Without an interval (for example /loop check the build): self-paced follow-ups chosen by the agent.
/loop list shows active jobs; /loop clear cancels all of them.

Install the loop skill

Download the ZIP from the skill's page on this directory and unzip it. Then place the folder in one of these locations, depending on how you want to use it:

You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.

Personal use across projects: unzip into ~/.claude/skills/
Single project only: unzip into your project's .claude/skills/ folder

Trigger the skill with natural requests

Start the input with /loop. The skill reads the input in this order: a bare /loop starts a self-paced autonomous loop; a leading interval token such as 5m makes it a fixed-interval loop; a trailing "every" clause with a time expression also sets an interval; anything else is a self-paced prompt.

An "every" phrase only counts as an interval when a time expression follows it. "check every PR" has no interval and runs as a self-paced prompt.

/loop check the build
/loop 5m check the build
/loop check the PR every 30m
/loop run tests every 5 minutes
/loop list
/loop clear

/loop 5m check the build

Understand intervals and rounding

Intervals are converted to cron expressions. Supported suffixes are s (seconds), m (minutes), h (hours) and d (days). Seconds are rounded up to the nearest minute because cron's minimum granularity is one minute.

If an interval does not divide its unit cleanly, the skill picks the nearest clean interval and tells you what it rounded to before scheduling. For example, 90m is 1.5 hours, which cron cannot express, so it is not scheduled exactly as written.

5m becomes */5 * * * * (every 5 minutes)
2h becomes 0 */2 * * * (every 2 hours)
1d becomes 0 0 */1 * * (every day at midnight local time)
30s is treated as 1 minute

Cancel and review jobs

Use /loop list to display the active jobs. Use /loop clear to cancel every job the skill returns; the agent reports how many were cancelled. You can also cancel one job with the agent's delete tool, using its job ID, which the confirmation message includes.

Recurring jobs created by the fixed-interval path expire automatically 7 days after creation by default, though the configured limit shown in the scheduling tool's description may differ.

/loop list: show jobs
/loop clear: cancel all jobs
Cancel one job: use the CronDelete tool with its job ID

/loop list

Check what the skill needs and its safety notes

The skill needs an agent runtime that provides the cron_create, cron_list, cron_delete and loop_wakeup tools, plus the loop task file reader and resolver TypeScript modules bundled with qwen-code. It runs scripts, and it does not use the network according to the skill's listing.

Each time a recurring job fires, the prompt runs again with the agent's tools. Treat a loop as something that acts repeatedly on your behalf, not a one-off check. Do not use it for tasks that must never repeat or run unattended.

Requires: cron_create, cron_list, cron_delete, loop_wakeup tools
Runs scripts: yes
Uses network: no
Scheduled jobs keep firing until cleared with /loop clear or the delete tool

Tips for reliable loops

Keep prompts specific so each tick does one clear check. Confirm the cron expression in the confirmation message matches the cadence you wanted. Run /loop list after starting a loop to make sure only the jobs you expect are active.

Do not expect exact-to-the-second timing. Intervals are rounded to whole minutes.

Write a prompt that states what to check and what to report.
Use /loop list regularly to confirm active jobs.
Clear jobs as soon as the watch is no longer needed.
Use a longer interval for things that change slowly, such as a deploy or PR review.

What people use it for

Watching a build and reporting when it finishes or fails with /loop 5m check the build.
Checking a pull request every 30 minutes with /loop check the PR every 30m.
Running a test command every few minutes during a long debugging session.
Sending a periodic status check on a deploy using a fixed interval.
Letting the agent pick its own check-in times for an open-ended task with /loop check the build.

Frequently asked questions

Is the loop agent skill free?

The skill is published under the Apache-2.0 licence, as listed on this page. Check the repository and licence file for the exact terms before reuse.

How do I install the loop skill?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or a project's .claude/skills/ folder for one project. You can also upload it in the Claude apps under Settings → Capabilities → Skills.

What does /loop 5m do?

It runs your prompt immediately and then creates a recurring job every 5 minutes, using the cron expression */5 * * * *. The job keeps firing until you cancel it.

How do I stop a loop?

Run /loop clear to cancel all jobs, or use the CronDelete tool with a specific job ID to cancel one. Use /loop list first to see what is running.

Why did my interval change?

Intervals are rounded to whole minutes, and values that cron cannot express, such as 90m, are changed to the nearest clean interval. The agent tells you what it rounded to before scheduling.

What is the difference between /loop 5m and /loop check the build?

/loop 5m creates a fixed, recurring schedule. /loop check the build with no interval runs once now and lets the agent choose when to check again, up to about an hour apart, stopping when the task is complete or blocked.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use the loop skill if you want an agent to keep checking a build, deploy or pull request and report back, and you are comfortable cancelling jobs with /loop clear. Pick something else if the task must never repeat, needs exact-to-the-second timing, or runs in an environment where you cannot easily list and cancel scheduled jobs.

Download the skill