roadmap
- 01
- 02
- 03
- 04
- 05
- 06
- 07
- 08
Talk through Git with confidence in interviews: how version control works under the hood and how you use it in team work. Covers core Git, pull requests and code review, Actions, Pages and security.
447 of 447 questions in your plan: 0 known, 0 in progress, 0 due, 0 skipped, 447 new. 0 cut, outside your plan.
Few loops schedule time for Git by name. It is tested in passing, inside rounds about something else: a screening question on what git fetch does that git pull does not, a few minutes of the behavioral conversation on how your team reviews and merges, a take-home whose commit history is read before the code. DevOps, platform and release-engineering loops are the exception, where the pipeline and the repository settings can fill a technical conversation.
The questions are usually scenarios. You ran git reset --hard on the wrong branch: is the work gone? A rebased branch had already been pushed and a colleague built on it; what does a force-push do to them, and what does --force-with-lease check first? An access key was committed and pushed an hour ago.
What separates candidates is whether they reason from the model or from remembered commands. Someone who knows a branch is a movable pointer says the committed work survived: the reset moved the pointer, the reflog records where it was, and only work that was never committed is at real risk. The stronger answer on the leaked key starts with revoking it, since a pushed secret must be treated as public, and leaves rewriting history for cleanup. Senior conversations add the platform: a pull_request_target workflow that checks out a fork's code while holding the repository's secrets, a GITHUB_TOKEN scoped wider than its job, and branch rules that keep main trustworthy without slowing each merge.
Commits as snapshots, and the graph that links them, come first, because the commands after them are explained as objects created and pointers moved. Staging is how the next snapshot gets built, merging and rebasing are two ways of joining lines of work, and the reflog is why so little is ever lost for good. The tooling around that core closes Git itself. Large file storage follows as the deliberate exception to the model: a small pointer file is committed where the content would have been.
Collaboration moves from your repository to the shared one. GitHub is organized around the pull request: branch rules decide what a merge requires, and review, CI checks and the security features are the things those rules can require. Code review as a practice follows the buttons and belongs to no one tool: what to look for in a change, and how to say it.
The plan then turns to automation. Actions has appeared inside the GitHub module as one platform feature; its own module reads a workflow file from triggers through runners to secrets and environments, where the trust questions sit. Packages and Pages follow as places a workflow publishes to.
The last two stretches are tools around the same repository. The GitHub CLI does from a script what the earlier modules did in a browser, down to raw API calls, and Codespaces moves the development machine into a devcontainer the repository defines. AI comes last because it leans on the rest: Copilot's cloud agent works on a branch and hands back a pull request that meets the same review and checks, and GitHub Models offers hosted models under the same account.
Most readers of this page are preparing for something else, and Git is a handful of questions inside it. For them the object model is worth more than a list of commands. That argues against the default study order: on by topics rather than by frequency, the queue opens on commits and pointers instead of whichever command question is asked most. Stop once the pull request and its branch rules feel solid. Keep a throwaway repository open as you go: a commit you lost on purpose and recovered from the reflog is an easy answer under follow-up. Candidates for DevOps, platform or release-engineering roles weigh the plan the other way: skim the commands and spend the time on Actions, the security features and the API, where the trust questions from the senior rounds live. Past its first stretch the plan is GitHub's platform, apart from the review practice. Merge requests, `.gitlab-ci.yml` pipelines and GitLab runners belong to the [GitLab CI/CD roadmap](/roadmaps/gitlab-ci); and if building pipelines is the whole job rather than part of it, [GitHub Actions](/roadmaps/github-actions) has a roadmap of its own.
Most readers of this page are preparing for something else, and Git is a handful of questions inside it. For them the object model is worth more than a list of commands. That argues against the default study order: on by topics rather than by frequency, the queue opens on commits and pointers instead of whichever command question is asked most. Stop once the pull request and its branch rules feel solid. Keep a throwaway repository open as you go: a commit you lost on purpose and recovered from the reflog is an easy answer under follow-up.
Candidates for DevOps, platform or release-engineering roles weigh the plan the other way: skim the commands and spend the time on Actions, the security features and the API, where the trust questions from the senior rounds live.
Past its first stretch the plan is GitHub's platform, apart from the review practice. Merge requests, .gitlab-ci.yml pipelines and GitLab runners belong to the GitLab CI/CD roadmap; and if building pipelines is the whole job rather than part of it, GitHub Actions has a roadmap of its own.