The spec your coding agent actually follows.

# AGENTS.md

Instructions for AI coding agents working on **Ledgerly** (saas).
Generated by prompt-generator-website.com. Same answers, same file, byte for byte.

## Where the truth lives

The full specification is reproduced below and also stored in `spec.md` at the repo root.
Those two are the same document. If code and spec disagree, the spec wins until a human
changes the spec — do not "fix" the spec to match code you just wrote.

## Working agreement

- Read the whole specification before touching a file. Not the first section, all of it.
- Build section by section, in the order given. Finish one before starting the next.
- The "Implementation rules" section at the end is a set of invariants, not suggestions.
- Do not invent tables, routes or dependencies that are absent from the spec. If something
  is genuinely missing, say so and stop; do not improvise it.
- Do not swap the tech stack for one you like better.
- Keep the file structure exactly as described.

## Full specification

# Complete Specification: Ledgerly

> This document is an exhaustive specification for building the **Ledgerly** project.
…

# CLAUDE.md — Ledgerly

The specification and the working agreement live in `AGENTS.md`, imported below, so Claude Code
and every AGENTS.md-aware agent work from the same brief. Edit `AGENTS.md`, not this file.

@AGENTS.md

## 3. Database Schema

### Table `users`
| Column | Type | Description |
|--------|------|-------------|
| id | INT PK AUTO | Unique identifier |
| email | VARCHAR(255) UNIQUE | Login email |
| password_hash | VARCHAR(255) | Hashed password (bcrypt) |
| first_name | VARCHAR(100) | First name |
| last_name | VARCHAR(100) | Last name |
| role | ENUM | User role |
| created_at | DATETIME | Creation date |
| updated_at | DATETIME | Last modified |

### Business-specific entities
Create tables for the following entities (with appropriate relationships):
Client, Invoice, LineItem, Expense, Receipt, PaymentReminder, Report

For each entity:
- Add `id`, `created_at`, `updated_at`
- Add appropriate foreign keys
- Add indexes on frequently queried columns

### Table `subscriptions`
| Column | Type | Description |
|--------|------|-------------|
| id | INT PK | Identifier |
| user_id | FK -> users | User |
| plan | VARCHAR | Subscribed plan |
| status | ENUM | active, canceled, expired |
…

## Data model

### `users`
Account identity for NextAuth (email/password + OAuth)

| Column | Type | Null | Notes |
|---|---|---|---|
| `id` | TEXT (cuid) PK | no | NextAuth compatible string id |
| `email` | VARCHAR(255) | no | UNIQUE, lowercased on write |
| `email_verified` | TIMESTAMPTZ | yes | NextAuth field |
| `password_hash` | VARCHAR(255) | yes | NULL for OAuth-only accounts; bcrypt cost 12 |
| `name` | VARCHAR(200) | yes |  |
| `image` | TEXT | yes | Avatar URL |
| `is_platform_admin` | BOOLEAN | no | DEFAULT false; gates /admin |
| `banned_at` | TIMESTAMPTZ | yes | Banned users cannot sign in |
| `locale` | VARCHAR(5) | no | DEFAULT 'en', one of en\|fr |
| `default_org_id` | TEXT | yes | Last used organization |
| `deleted_at` | TIMESTAMPTZ | yes | GDPR soft delete before purge job |
| `created_at` | TIMESTAMPTZ | no | DEFAULT now() |
| `updated_at` | TIMESTAMPTZ | no |  |

Relations: default_org_id -> organizations.id, set null
Indexes: UNIQUE(email); INDEX(banned_at)

### `accounts`
NextAuth OAuth provider links (Google, GitHub)

| Column | Type | Null | Notes |
|---|---|---|---|
| `id` | TEXT PK | no |  |
…