This epic tracks **Phase 1 - Library extraction** of the [GitLab Rails Decomposition](https://gitlab.com/groups/gitlab-org/-/work_items/22372): pulling cross-cutting, ActiveRecord-free platform code out of `lib/gitlab/` and `ee/lib/gitlab/` into independently testable in-repo gems under `gems/`. ## Why - **Faster CI** - each gem ships its own isolated test suite that runs only when the gem changes, instead of riding the full monolith suite. - **Clearer ownership** - every gem maps to a `feature_category:`, making team boundaries explicit. - **Enforced isolation** - hidden dependencies become visible CI failures (a RuboCop cop blocks references to a gem's `Internal` namespace from outside the gem) instead of implicit coupling. - **Agent-ready boundaries** - explicit contracts and self-contained context let AI agents work autonomously within a single gem. ## Approach Sequencing is **leaves-first / common-infrastructure-first**: shared utilities are extracted before domain gems, so each extraction removes stub categories for the next. ActiveRecord touchpoints are replaced with configuration callbacks wired from the monolith, so gems stay AR-free and runnable standalone. Each gem under this epic is tracked by its own child issue, which states the gem being created, what it covers, and what stays in the monolith. ## References - Handbook: [Extracting cross-cutting libraries into gems](https://gitlab.com/gitlab-com/content-sites/handbook/-/merge_requests/20064) (ADR-004) - R&D strategy docs: [`docs/strategy.md`](https://gitlab.com/ayufan-repos/gitlab-r-and-d/poc-gitlab-rails-agent-harness/-/blob/main/docs/strategy.md) and [`docs/lib/extraction.md`](https://gitlab.com/ayufan-repos/gitlab-r-and-d/poc-gitlab-rails-agent-harness/-/blob/main/docs/lib/extraction.md) (`poc-gitlab-rails-agent-harness`) --- *Claude helped with this epic.*