Schema Designer

schema-designer

Design a Postgres schema that's correct first, fast second.

Process

1. List the entities and relationships before writing SQL. State cardinality (1:1, 1:N, N:M).
2. Normalize to 3NF unless there's a measured reason to denormalize. Note any denormalization explicitly.
3. Write the DDL following the conventions below.
4. Add indexes for every foreign key and every column used in frequent WHERE/ORDER BY. Justify each.
5. Add constraints — NOT NULL, CHECK, UNIQUE, FK ON DELETE behavior. Don't leave integrity to app code.

Conventions

• Tables: plural snakecase (order_items). Columns: snakecase.
• Primary key: id uuid primary key default gen_random_uuid() (Supabase-friendly).
• Timestamps: created_at timestamptz not null default now(), updated_at timestamptz (with a trigger).
• Foreign keys: _id, always with an explicit references + on delete rule.
• Money: numeric(12,2), never float. Booleans: is_/has_ prefix.

Output

• The create table statements.
• The index list with a one-line reason each.
• A note on which tables need RLS (hand off to rls-policy) and which columns hold PII.

Guardrails

• Don't ship a table holding user data with RLS unconsidered.
• Don't use text for everything — pick real types and constraints.
• For N:M, create the explicit join table; don't fake it with arrays unless justified.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: m-binimran
• Source: m-binimran/dev-pack
• License: MIT

Install and usage instructions live in the source repository linked above.