Motir

Estimate: 28m

Add the persistence for story points + the project estimation config as ONE Prisma migration, modelled so prisma migrate dev is drift-free (a second run reports "No difference detected").

WorkItem addition: storyPoints Decimal? @db.Decimal(6, 2) — the agile estimate, nullable (null = unestimated), camelCase (NO @map, matching work_item's existing camelCase columns — the same within-table-consistency call the schema records for sprintId / backlogRank; decision-ladder rung 2). Decimal, not Float — story points allow 0.5 increments (Jira-faithful) AND roll up via SUM; Decimal avoids floating-point drift in the roll-up sums (a real-product correctness concern), at the small fixed precision (6, 2). SEPARATE from the existing estimateMinutes (TIME) — both columns coexist; the project statistic config picks which one is THE planning estimate. Index: add storyPoints to a covering index ONLY if the roll-up aggregates need it (the sprint roll-up filters on the existing (projectId, sprintId, backlogRank) composite; the epic roll-up walks parentId — measure, do not add a speculative index).

Project additions (the estimation config — PROJECT-scoped per the module header):

• estimationStatistic EstimationStatistic @default(story_points) — which value is THE planning estimate the surfaces display + the roll-ups sum.
• pointScale PointScale @default(fibonacci) — the suggested-deck the estimate picker offers.
• customScaleValues Float[] @default([]) — the project-defined deck values, used ONLY when pointScale = custom (these are UI SUGGESTIONS — summation uses the Decimal storyPoints, so Float[] for the suggestion list is fine and avoids a Decimal-array column). All camelCase, no @map, matching project's existing workflowPolicyMode / accessLevel columns. They backfill to the defaults on migration (every existing project becomes Story Points + Fibonacci — the agile default, no lock-out).

Enums:

• EstimationStatistic (@@map("estimation_statistic")): story_points, time_estimate, issue_count — the three Jira estimation statistics. All values exist now so later stories add no enum ALTER.
• PointScale (@@map("point_scale")): fibonacci, linear, custom.

No new FK (the columns are scalars / a scalar list on existing tables), so the FK-as-@relation rule is n/a here — but the migration must STILL be drift-free (no spurious change on a second migrate dev), and no other model changes.

Acceptance criteria

• schema.prisma gains work_item.storyPoints Decimal? @db.Decimal(6, 2) (camelCase, nullable, separate from estimateMinutes) and the project columns estimationStatistic / pointScale / customScaleValues with the defaults above, plus the EstimationStatistic + PointScale enums (@@map-ed).
• One migration add_story_points_and_estimation_config creates the column, the project columns, and the two enums; existing projects backfill to story_points + fibonacci + []. pnpm prisma migrate dev applies cleanly and a SECOND run reports "No difference detected" (no drift).
• pnpm prisma generate + pnpm typecheck + pnpm build pass; no other model changes; no speculative index (add one only if a roll-up query measurably needs it).

Context refs

• prisma/schema.prisma model WorkItem (the estimateMinutes / sprintId / backlogRank columns + the camelCase-no-@map convention comment) and model Project (workflowPolicyMode / accessLevel — the project-config-column pattern to mirror)
• motir-core/CLAUDE.md (the migration drift rule — even without an FK, a second migrate dev must report "No difference detected") + the bug-attachment-fk-migration-drift precedent
• Jira estimation statistics (Story Points / Original Time Estimate / Issue Count) + planning-poker decks (Fibonacci / linear) as the mirror for the enums