## Problem Historically, rather than using our `run_command` tool a separate tool has been used to run git commands called `run_git`. This was originally necessary to provide more control over git functionality, and it sets up the Git environment for usage by the agent. However, we see repeated and hard to debug problems (e.g. https://gitlab.com/gitlab-com/request-for-help/-/work_items/4013#note_3017217597, https://gitlab.com/gitlab-org/editor-extensions/gitlab-lsp/-/issues/1897) with the Git functionality when flows execute on runners. However, this is persistently creating problems: 1. Git configuration changes need to be funneled through to the LSP (see https://gitlab.com/gitlab-org/gitlab/-/work_items/587996), to change the Git command accordingly. 2. If an agent chooses to use the normal `run_command` to run something git specific it might easily fail because the setup in the image is not correct. 3. `run_git_command` might show weird behavior if used locally as it expects the environment of a runner rather than the local one. ## Desired Outcome Git environment in GitLab runner execution is prepared in a way so that any git command being run works as expected even if it is run directly via `run_command`, e.g. `run_command('git push')` should push to the correct origin with the correct user.