Seed Data Generator

Purpose

Generate idempotent database seed scripts with realistic test data.

When to Use (STRICT CRITERIA)

Only invoke this skill when ALL of these are true:

1. User explicitly requests seed/test data generation with phrases like:
   • "generate seed data"
   • "create test data script"
   • "populate database with mock data"
   • "need sample data for testing"
2. User specifies schema/models that need data
3. NOT for simple CRUD code or component generation

When NOT to Use

• Regular feature development
• Component/API code generation
• Bug fixes or refactoring
• User asks general "how to" questions
• Unclear requirements

Usage Pattern

User triggers with explicit command: "Generate a seed script for [Product, User, Order] models with 50 realistic entries each"

Output Format

1. Idempotent Script Template (TypeScript)

// scripts/seed-[entity].ts
import { PrismaClient } from '@prisma/client'
// or: import payload from 'payload'
// or: WordPress REST API client

const prisma = new PrismaClient()

async function seed[Entity]() {
  console.log('Seeding [entity] data...')

  const data = [
    // Realistic data array (50-100 items)
  ]

  for (const item of data) {
    await prisma.[entity].upsert({
      where: { id: item.id }, // or slug/email
      create: item,
      update: item,
    })
  }

  console.log(`Seeded ${data.length} [entity] records`)
}

seed[Entity]()
  .catch(console.error)
  .finally(() => prisma.$disconnect())

2. Realistic Data Generation

• Use varied, production-like data
• Include edge cases (long strings, special chars, nulls)
• Realistic dates (past year range)
• Proper relationships (foreign keys)

3. Make it Idempotent

Always use:

• upsert (Prisma)
• findOneAndUpdate with upsert: true (Mongoose)
• INSERT ... ON CONFLICT DO UPDATE (raw SQL)
• wp.posts().id(123).update() with existence check (WordPress)

4. Include Run Instructions

// package.json
{
  "scripts": {
    "seed": "tsx scripts/seed-all.ts",
    "seed:products": "tsx scripts/seed-products.ts"
  }
}

Token Efficiency Rules

1. Generate data inline (no external files)
2. Use concise but realistic content
3. Batch related entities in one script
4. Ask clarifying questions FIRST if unclear:
   • "How many records per entity?"
   • "Which entities need seeding?"
   • "What ORM/database?"

Example Invocation

User says: "I need test data for my e-commerce app - products, categories, and users"

Claude asks: "I can generate seed scripts for those. Quick questions:

1. How many of each? (default: 50 products, 10 categories, 20 users)
2. Are you using Prisma, Payload, or WordPress?
3. Any specific requirements (price ranges, user roles, etc.)?"

Then generates: Complete seed scripts after confirmation.

Anti-Patterns (DO NOT DO)

• Generate seed data when user just asks for a component
• Auto-suggest seeding for every database model
• Create seed data without schema context
• Generate when user says "create a Product model" (wait for explicit seed request)

Success Criteria

• User runs script multiple times safely
• Data looks production-ready
• Script handles existing data gracefully