> ## Documentation Index > Fetch the complete documentation index at: https://docs-stg.outerlayer.ai/llms.txt > Use this file to discover all available pages before exploring further. # Schedule automations > Read a repository at set times with an agent, and file issues only when it finds work. A schedule automation starts a run at each time you set. In a run, an agent reads one repository with your instructions and reports what it finds. OuterLayer files an issue for each thing the run reports. A run that finds nothing files nothing. Each issue shows on the Work page as **Not started** until a member chooses **Build** on the item's page. See [Build an item that is not started](https://docs-stg.outerlayer.ai/the-work-page.md#build-an-item-that-is-not-started). For example, a run each night can look for tests that fail now and then, and file one issue for each. ## What a run can do A run reads. It does not change anything. * **It reads one repository.** Its token covers only the repository you choose. The token can read the files, the workflow runs and the checks. It has no write permission. * **It cannot push, open a pull request or write to a tracker.** It holds no credential that can write. * **It gets no build variables and no destinations.** See [Destinations and secrets](https://docs-stg.outerlayer.ai/destinations-and-secrets.md). * **It never starts a build.** The agent reads text from the repository, and that text can try to steer the agent. So a member reads each issue first. See [Why an issue from outside the team waits](https://docs-stg.outerlayer.ai/automations.md#why-an-issue-from-outside-the-team-waits). The agent reports what it found, and OuterLayer files the issues after the run ends. OuterLayer files them only in the repository the run read, or in a Linear team or Jira project of your factory. It removes images from the issue body and writes each mention as plain text. ## Before you start * You need the **Edit factory settings** permission. * The factory needs a connected GitHub repository. See [Connect a repository](https://docs-stg.outerlayer.ai/connect-repository.md). * The factory needs a self-hosted runner whose builds run in a microVM or a container. See [Container builds](https://docs-stg.outerlayer.ai/container-builds.md). A runner takes a run only when all of these are true: * It has sent a heartbeat within three of its own poll intervals. Otherwise OuterLayer counts it as not running. * Its builds do not run with `shared-user` isolation. * It is not a managed runner. A managed runner takes no scheduled run. * OuterLayer does not provide its sandboxes. A runner whose `runner.sandbox.provider` setting is `"outerlayer"` takes no scheduled run. * It has a command for a run. * Its `runner.queues` setting includes `implement`. See the [config file](https://docs-stg.outerlayer.ai/reference/config-file.md). * Its `repos` filter lets the automation's repository through. * It is up to date. See [Update a runner](https://docs-stg.outerlayer.ai/update-a-runner.md). A runner that meets all of these can still be unavailable for a time. It can be draining, restarting, holding for an update, blocked or low on disk. It takes no run until that state ends. A runner takes builds before it takes runs, so a run can wait while the runner is busy. When no runner can take the runs of the repository you choose, the form and the automation's page show a warning. The warning names the cause and links to this list. It does not stop you from saving the automation or choosing **Run now**. If your factory builds on managed runners, the warning also says that a managed runner takes no scheduled run. ## Add a schedule automation 1. Open **Settings**, then **Automations**. 2. Choose **New automation**, then choose **Schedule**. 3. Enter a name. 4. Under **Schedule**, choose **Hourly**, **Daily**, **Weekdays** or **Weekly**. **Hourly** runs at the start of each hour. 5. Set the **Time**, and the **Day** for **Weekly**. 6. Choose the **Time zone**. The form starts with the time zone of your browser. 7. Choose the **Repository**. The run reads its default branch. 8. In **What to look for**, tell the agent what to look for. The field holds up to 2,000 characters. 9. Under **File issues in**, choose the same repository, or a Linear team or Jira project of the factory. 10. Set the **Daily limit**: the most issues the automation files in 24 hours. The default is 5, and you can set 1 to 100. 11. Choose **Save**. The automation is on as soon as you save it. It shows in the **Schedules** group. Its page shows its settings and the time of the next run. A due time uses the local time of the time zone. A daily run at 09:00 in New York stays at 09:00 New York time when the clocks change. When the last run is still pending or running at a due time, OuterLayer starts no run for that time. The run log shows **Skipped**. ## Start a run now Choose **Run now** on the automation's page. You need the **Edit factory settings** permission. **Run now** works while the automation is off. The button is off while the last run is pending or running. It works again after a run fails because no runner took it in 7 days. ## How a run files issues The agent gives each issue a topic: short text that names what the issue is about. OuterLayer uses the topic and the daily limit to hold issues back: * **One open issue for each topic.** While the issue that the automation filed for a topic is open, OuterLayer files no other issue for that topic. * **A closed topic stays closed.** A member can close an issue with no merged pull request linked to it. Then OuterLayer files no new issue for that topic. The same applies when a member chooses **Remove from Work** on the issue's item. If a pull request merged before the issue closed, the fix shipped. A later run that finds the topic again files a new issue. * **The daily limit.** The automation files at most its daily limit of issues in 24 hours. At the limit, the agent stops reporting. A report for a topic that OuterLayer holds back does not use up the limit. The agent can still report new problems, and the run log shows that the run found the topic again. * **Secrets.** OuterLayer files nothing for an issue whose title, topic or body looks like it holds a secret, such as a token. OuterLayer remembers each topic for as long as the automation exists. A topic matches an earlier topic only when its text is the same. A topic that the agent words differently is a new topic. The run log names each issue that OuterLayer did not file, and why. ### Topics that stay closed The automation's page has a section **Topics that stay closed**. It lists each topic whose issue a member closed without a fix, with a link to that issue. The section is not there when no topic is closed. To let OuterLayer file a topic again, choose **File again** on its row. You need the **Edit factory settings** permission. **File again** files no issue at once. OuterLayer files a new issue the next time a run finds the topic. A run stops after 30 minutes. The run log then shows **Failed**, and the run files nothing. The agent reads its topics and reports with three CLI commands that only a run's agent can use. See [Run commands](https://docs-stg.outerlayer.ai/reference/cli-run.md). ## Read the run log The automation's page shows the run log. Each row has the time, who started the run, the result, the cost and a link to the session of the run. OuterLayer keeps each row for 90 days. | Result | What it means | | - | - | | **Pending** | No runner has taken the run yet. The row says why. See the reasons after this table. A pending run keeps waiting, and a runner that comes back takes it. | | **Running** | A runner is reading the repository. | | **Found** | The run reported issues. The row links to each filed issue, and names each issue that OuterLayer did not file. | | **Nothing to do** | The run found nothing. The row shows the reason the agent gave. | | **Skipped** | The last run was still pending or running at the due time. | | **Failed** | The run stopped at its time limit, or it ended without a result. It also means that no runner took the run in 7 days. The row then gives the reason that held at that time. | A pending row gives one of four reasons: * **No runner of the factory is running.** No runner has sent a heartbeat within three of its poll intervals. * **No running runner takes this automation's runs.** A runner is running, but none meets the conditions in [Before you start](#before-you-start) for this repository. * **Every runner that takes this run is unavailable.** The row names each runner and its state, such as "ollie is low on disk". * **The run waits for a free runner.** A runner takes builds first, so the run waits until a runner is free. A run that waits 7 days ends as **Failed**. OuterLayer then switches the automation off. The row says that no runner took the run in 7 days. It then gives the reason in the past tense, such as "because no runner of the factory was running". When a tracker refuses an issue, the row shows the tracker's answer. Sometimes OuterLayer cannot tell whether the tracker created an issue. The row then says where to look, and OuterLayer does not file the issue again. ## Notices and the switch-off after repeated failures OuterLayer tells the creator of a schedule automation when a run waits for a runner or fails. It sends the same notices for other automations that a member adds. See [Notices about an automation](https://docs-stg.outerlayer.ai/automations.md#notices-about-an-automation) for who gets them and what each email says. A schedule automation that keeps failing is a waste of runs. So OuterLayer switches it off in either of these cases: * Five runs in a row fail. * A run waits 7 days and no runner takes it. This ends the run as **Failed**. A switched-off automation has no next run. It shows as **Switched off** in the Inbox, and its page says why and when. OuterLayer sends one email for the switch-off. To turn the automation on again, use the switch on its page. A run then starts at the next due time. To leave the automation off, choose **Keep it off** on its page. The notice then goes away, and the automation stays off. A save by a member also sets the count of failed runs in a row to zero. OuterLayer never switches off a webhook automation or an automation for issues. To change, switch off or delete a schedule automation, see [Change, switch off or delete an automation](https://docs-stg.outerlayer.ai/automations.md#change-switch-off-or-delete-an-automation). This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.