database-schema-designer

Add an organization_id column to every tenant-scoped table and enforce isolation with Row-Level Security policies that filter rows by the current user's organization membership. Combine this with composite indexes on organization_id plus status for query performance.

With Drizzle, use $inferSelect and $inferInsert on table definitions to derive TypeScript types automatically. With Prisma, the generated client provides model types, and this Skill produces equivalent interfaces or Pydantic models for Python.

Prisma offers declarative schemas with a generated client, Drizzle provides SQL-like TypeScript schema definitions with lightweight migrations, and Alembic handles Python/SQLAlchemy migrations with explicit upgrade and downgrade functions. The Skill generates examples for all three.

RLS enforces tenant isolation at the database layer, so even buggy application queries cannot leak rows across organizations. It complements rather than fully replaces application authorization, and policies must be tested with a non-superuser role.

Queries filtering WHERE deleted_at IS NULL perform full scans without a supporting index. Create a partial index on the filtered columns with the condition deleted_at IS NULL so active-row lookups stay fast.

Use soft deletes with a deleted_at timestamp for auditable or recoverable data such as tasks, projects, and user records. Hard deletes suit ephemeral or compliance-purged data where retention is prohibited.