The Shell executor is a simple executor that allows you to execute builds locally to the machine that the Runner is installed. It supports all systems on which the Runner can be installed. That means that it’s possible to use scripts generated for Bash, Windows PowerShell and Windows Batch (deprecated).

Note: Always use the latest version of Git available. Additionally, GitLab Runner will use the git lfs command if Git LFS is installed on the machine, so ensure Git LFS is up-to-date when GitLab Runner will run using the shell executor.


The scripts can be run as unprivileged user if the --user is added to the gitlab-runner run command. This feature is only supported by Bash.

The source project is checked out to: <working-directory>/builds/<short-token>/<concurrent-id>/<namespace>/<project-name>.

The caches for project are stored in <working-directory>/cache/<namespace>/<project-name>.


  • <working-directory> is the value of --working-directory as passed to the gitlab-runner run command or the current directory where the Runner is running
  • <short-token> is a shortened version of the Runner’s token (first 8 letters)
  • <concurrent-id> is a unique number, identifying the local job ID on the particular Runner in context of the project
  • <namespace> is the namespace where the project is stored on GitLab
  • <project-name> is the name of the project as it is stored on GitLab

To overwrite the <working-directory>/builds and <working-directory/cache specify the builds_dir and cache_dir options under the [[runners]] section in config.toml.

Running as unprivileged user

If GitLab Runner is installed on Linux from the official .deb or .rpm packages, the installer will try to use the gitlab_ci_multi_runner user if found. If it is not found, it will create a gitlab-runner user and use this instead.

All shell builds will be then executed as either the gitlab-runner or gitlab_ci_multi_runner user.

In some testing scenarios, your builds may need to access some privileged resources, like Docker Engine or VirtualBox. In that case you need to add the gitlab-runner user to the respective group:

usermod -aG docker gitlab-runner
usermod -aG vboxusers gitlab-runner

Selecting your shell

GitLab Runner supports certain shells. To select a shell, specify it in your config.toml file. For example:

  name = "shell executor runner"
  executor = "shell"
  shell = "powershell"


Generally it’s unsafe to run tests with shell executors. The jobs are run with the user’s permissions (gitlab-runner) and can “steal” code from other projects that are run on this server. Use it only for running builds on a server you trust and own.

Terminating and killing processes

The shell executor starts the script for each job in a new process. On UNIX systems, it sets the main process as a process group.

GitLab Runner terminates processes when:

GitLab 13.0 and earlier

On UNIX systems gitlab-runner sends a SIGKILL to the process to terminate it, because the child processes belong to the same process group the signal is also sent to them. Windows sends a taskkill /F /T.

GitLab 13.1 and later

On UNIX system gitlab-runner sends SIGTERM to the process and its child processes, and after 10 minutes sends SIGKILL. This allows for graceful termination for the process. Windows don’t have a SIGTERM equivalent, so the kill process is sent twice. The second is sent after 10 minutes.

If for some reason this new termination process has problems with your scripts but works with the old method you can set the feature flag FF_SHELL_EXECUTOR_USE_LEGACY_PROCESS_KILL to true, and it will use the old method. Keep in mind that this feature flag will be removed in GitLab Runner 14.0 so you still need to fix your script to handle the new termination.