- Home
- Skills
- Testing & Debugging
- Realistic Seed and Test Data Generator for Any Schema
Works with the AI Agents you already use
Realistic Seed and Test Data Generator for Any Schema
Turns your schema into a deterministic, constraint-correct seed script with edge cases, safe fake contacts, reset and validation. Verdict READY, GAPS or UNSAFE.
$5
Realistic Seed and Test Data Generator for Any Schema
Example session with this skill installed
Here's my prisma/schema.prisma (User, Product, Order, OrderItem on Postgres). Write a demo-size seed with realistic data and review my current prisma/seed.ts too.
- Read your context and instructions
- Compiled the realistic seed
Seed plan — shop-api · PostgreSQL · profile: demo
Verdict
GAPS — new seed is READY; your existing prisma/seed.ts is GAPS (no reset, Date.now() dates, default faker emails)
Assumptions: seed 42 · ref date 2026-01-01Z · en_US + 10% ja_JP/de_DE/ar_EG · script not executed
Insert order
- User (200 + 6 edge) → 2. Product (50) → 3. Order → 4. OrderItem · Cycles: none
Constraints handled
| Table.column | Constraint | How generated |
|---|---|---|
| User.email | unique | first.last.<i>@example.com |
| User.phone | nullable | +1 312 555 01XX, 10% NULL |
| Order.status | enum | weighted, declared values only |
| OrderItem(orderId, productId) | composite unique | sample products without replacement per order |
| OrderItem.quantity | CHECK > 0 | 1–3 mostly, max 12 |
Distributions (modelling choices)
- Orders per user: 0 for 30%, 1–2 for 45%, 3–10 for 20%, 11–50 for 5%
- Status: DELIVERED 70 / SHIPPED 12 / PENDING 10 / CANCELLED 6 / REFUNDED 2
- createdAt: last 18 months before ref date, weighted to recent months
Edge-case rows
- edge-1:
山田 太郎, ja_JP · edge-2:محمد عبد الله - edge-3: name at VARCHAR(100) limit · edge-4: all nullable fields NULL
- edge-5: user with 0 orders · edge-6: order placed 2024-02-29T23:59:59Z
Files
// prisma/seed.ts (excerpt)
import { faker } from '@faker-js/faker';
faker.seed(42);
faker.setDefaultRefDate('2026-01-01T00:00:00Z');
const status = () => faker.helpers.weightedArrayElement([
{ weight: 70, value: 'DELIVERED' }, { weight: 12, value: 'SHIPPED' },
{ weight: 10, value: 'PENDING' }, { weight: 6, value: 'CANCELLED' },
{ weight: 2, value: 'REFUNDED' }]);
Run: npx prisma db seed -- --reset (refuses if DATABASE_URL host is not localhost)
Validation (expected)
- User = 206, Product = 50; orphan OrderItem → Order = 0
- duplicate emails = 0; md5 of User identical across two runs
Findings on existing seed.ts
- [GAP] line 12
createdAt: new Date()→ seeded refDate - [GAP] no reset → add scoped TRUNCATE ... RESTART IDENTITY CASCADE
- [GAP]
faker.internet.email()→ reserved example.com domain
Connects securely to your tools. The creator never sees your data.
About this skill
Most seed scripts fail in the same ways: inserts break on foreign keys or unique constraints, every run produces different data, all rows look identical, and sooner or later someone pastes in real customer emails or phone numbers. Asking an AI for "some fake data" usually gives you uniform rows, wall-clock dates and Faker emails on real mail domains.
This skill reads your schema and writes a seed you can commit.
- Parses SQL DDL, Prisma, Drizzle, Django, Rails, SQLAlchemy, TypeORM, JSON Schema, OpenAPI or a CSV header
- Orders inserts by foreign key, handles cycles and self-references
- Satisfies unique, composite unique, enum, CHECK, length and NOT NULL rules by construction
- Fixed random seed and fixed reference date so runs are reproducible
- Skewed, realistic distributions (long-tail order counts, weighted statuses, recent-heavy dates, mixed timezones)
- A tagged edge-case set: unicode and RTL names, max-length strings, all-NULL rows, leap-day and boundary values
- Only reserved example domains, fictional NANP 555-01XX and Ofcom drama numbers, documentation IPs and published Stripe test cards
- Small, demo and load profiles, with batched inserts or COPY for large volumes
- Scoped, production-guarded reset plus validation queries (row counts, orphan FKs, duplicates, determinism hash, PII scan)
- Reviews an existing seed script and grades it READY, GAPS or UNSAFE
Who it is for: backend and full-stack developers, QA engineers and anyone preparing demo environments.
Honesty note: the distribution weights are illustrative defaults, not industry statistics. Library versions and Prisma seeding behaviour were checked against official docs in October 2026 and can change; the skill re-checks when web access is available and never reports row counts for a script it did not run.
The skill writes and reviews seed code; it does not anonymise production dumps. It relies on the schema you provide and may miss rules enforced only in application code. Distribution weights are modelling defaults, not measured data. Fictional phone ranges are verified for North America and the UK only. Library versions change, so pin and re-check them.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago
Requires
Permissions
No special permissions declared or detected