Ensign Scope

Title

Tags

Body (markdown)

How Ensign writes code. Generated scopes should assume these as defaults.

## TypeScript & code quality

- **Strict mode**, always. Explicit types on exported/public APIs; inference is fine internally. No `any` without a comment explaining why.
- Small, focused changes. Match existing patterns in the repo before deviating.
- Run typecheck + lint after substantive changes; don't mark work done on compile alone — run it.

## Database (TypeORM / Drizzle)

- **UUID primary keys** for all entities.
- **snake_case** column naming (TypeORM custom naming strategy).
- **Soft deletes everywhere** — every entity has `deleted_at: Date | null`; hard deletes are never used in business operations. All queries filter `deleted_at IS NULL`.
- Standard audit columns: `created_at`, `updated_at` (+ `deleted_at`).
- **Multi-tenancy**: every business entity has an indexed `company_id`; it is auto-injected, never trusted from the client.
- Monetary values: `decimal(40, 2)`; VAT/percentages: `decimal(5, 2)`.
- Index `company_id`, foreign keys, and frequently-queried columns.

## Migrations

- **Never delete a committed migration.**
- Generate migrations from entity changes (don't hand-edit schema drift); a pre-commit hook blocks entity changes without an accompanying migration.
- New non-null columns need a default or a backfill — no unguarded `NOT NULL` adds.

## Git & process

- **Conventional Commits** (`feat:`, `fix:`, `chore:`…).
- Never list AI tools as a contributor/co-author.
- UK English in prose and user-facing copy; conventional (US) spelling in code identifiers.
- Commit and push as work progresses.

Edit context doc