## Summary Customer Zero, the first stage of [MR Risk Assessment](https://gitlab.com/groups/gitlab-org/-/work_items/21610) (see the [phased implementation proposal](https://gitlab.com/groups/gitlab-org/-/work_items/21610#note_3647849128)). Ships a read-only, shadow-mode risk assessment for every eligible merge request: a deterministic score, a derived risk level, matched domain tags, and a plain-language rationale, computed once, at MR open or ready-for-review time, in parallel with base Duo Code Review. Nothing is gated on the result yet, and there is no way for a human to correct an assessment yet: break-glass overrides are deferred to [Beta](https://gitlab.com/groups/gitlab-org/-/work_items/23654), so this stage can ship the smallest possible foundation first. Terminology for this stage follows the [dictionary in the parent epic](https://gitlab.com/groups/gitlab-org/-/work_items/21610). "Tier" is no longer used in user-facing copy; the categorical output is a **risk level** (Low / Medium / High / Critical). ## Deliverable Every eligible MR gets a computed, visible risk assessment. No merge blocking, no required extra approvals, no re-scoring on push. ## Scope - Data model for assessments and outcomes (**not** overrides, see [Beta](https://gitlab.com/groups/gitlab-org/-/work_items/23654)) - Deterministic scoring: a catalog of risk domains, diff-based signals, a versioned scoring function - A `risk_classification/v1` Duo Agent Platform flow (gitlab-ai-gateway) that answers a fixed set of categorical claims about a change, with evidence, and never emits a score or confidence value itself - A REST write path for the flow to submit its claims, and a GraphQL read path for the merge request widget - Feature-flagged, license-gated rollout, wired into MR create and draft-to-ready - Frontend surfacing on the merge request ## Explicitly out of scope for this stage - Break-glass override of a risk level (see [Beta](https://gitlab.com/groups/gitlab-org/-/work_items/23654)) - Any per-namespace configuration of domains or thresholds (see [GA](https://gitlab.com/groups/gitlab-org/-/work_items/23655)) - Any consumer other than the merge request view itself (see [Tiered Review Routing](https://gitlab.com/groups/gitlab-org/-/work_items/23591)) - Outcome signals beyond what this stage's data model can hold (collected from [Beta](https://gitlab.com/groups/gitlab-org/-/work_items/23654) onward, reported in [GA](https://gitlab.com/groups/gitlab-org/-/work_items/23655)) ## Trigger and widget: resolved Both open questions from the original plan are settled now. The trigger uses the flow trigger platform. It fires when a merge request opens, and again when a draft is marked ready. There is no manual "run" button. The widget is the full risk widget. It is not a lighter affordance limited to low-risk MRs. ## Current state: not complete This stage is not complete. On `master`, an assessment stops at `queued`. The claims from the flow are stored, but nothing turns them into a score yet. No merge request shows a risk level today. The scoring library is built in full, but nothing in the running code calls it. [Issue 609292](https://gitlab.com/gitlab-org/gitlab/-/issues/609292) tracks this gap. [Merge request 253762](https://gitlab.com/gitlab-org/gitlab/-/merge_requests/253762) closes it: it adds the worker that calculates the score and completes the state machine. What already ships and works: - The data model: the `merge_requests_risk_assessments` and `merge_requests_risk_outcomes` tables. - The `risk_classification/v1` flow, which reads the diff and answers one claim per domain, with evidence. - The write endpoint the flow uses to submit its claims. - The GraphQL read path, gated by `Project#risk_classification_dap_available?`. - The merge request widget, which today shows a single queued notice until scoring lands.