メニュー
Ensure test seed data never causes unique constraint violations — UUID-based, timestamp-based, counter-based, or Faker patterns for any field with a UNIQUE index, primary key, or slug constraint.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
SOC 職業分類に基づく
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/lukemcqueen/hermes-cortex --skill test-seed-uniquenessコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
SKILL.md を表示中
Cross-server agent health monitoring using binary status vectors — deploy health endpoints on each agent, poll from orchestrator, alert on state transitions.
Wire a self-hosted Langfuse instance to Hermes Agent — generate API keys, configure env vars, enable the bundled plugin, install SDK, and verify traces flow.
Use before enforcement code changes or shared-repo commits.
| name | test-seed-uniqueness |
| version | 1.0.0 |
| category | software-development |
| description | Ensure test seed data never causes unique constraint violations — UUID-based, timestamp-based, counter-based, or Faker patterns for any field with a UNIQUE index, primary key, or slug constraint. |
| author | Hermes Agent |
| license | MIT |
| platforms | ["linux","macos","windows"] |
| metadata | {"hermes":{"tags":["testing","seed-data","factories","fixtures","database","uniqueness","pytest","factory-boy"],"related_skills":["change-test-loop","subagent-driven-development","root-cause-debugging"]}} |
Every test needs seed data. Agents instinctively write:
user = User.objects.create(email="alice@example.com", username="alice")
This works once. On the next test run, or when tests run in parallel, or when a second test also creates an alice:
django.db.utils.IntegrityError: duplicate key value violates unique constraint
The agent didn't do anything wrong — it just picked a natural-looking value for a field that happens to have a UNIQUE constraint. The database doesn't care how pretty the data is; it cares about uniqueness.
Every value inserted into a uniquely-constrained field MUST be dynamically unique at runtime.
Hardcoded strings in seed data are only safe for:
display_name, bio, age)status="active", role="admin")Everything else — emails, usernames, slugs, external IDs, phone numbers, API keys, invite codes, profile handles — gets the dynamic treatment.
When you see any of these in test output, stop and fix the seed data:
| Error Pattern | Likely Source |
|---|---|
duplicate key value violates unique constraint | PostgreSQL |
UNIQUE constraint failed | SQLite |
E11000 duplicate key error | MongoDB |
Duplicate entry '...' for key | MySQL |
unique constraint "..." violated | Generic ORM |
already exists. | Rails / Active Record |
IntegrityError: (1062, "Duplicate entry | MySQL via ORM |
First debugging step when seeing these: Look at the test's seed data or fixtures. Find the field whose value is hardcoded. Make it dynamic.
import uuid
def test_create_user():
user = User.objects.create(
id=uuid.uuid4(),
email=f"user-{uuid.uuid4()}@example.com",
)
Use when: The value is never shown to a human in tests. IDs, tokens, API keys.
from datetime import datetime
slug = f"test-project-{datetime.now().strftime('%Y%m%d%H%M%S%f')}"
Use when: You want roughly ordered test data, or the field appears in logs/output.
from itertools import count
_user_counter = count(1)
def make_user(email=None):
n = next(_user_counter)
return User.objects.create(
email=email or f"user{n}@example.com",
username=f"user{n}",
)
Use when: Creating multiple objects in a loop or parameterized test.
import secrets
suffix = secrets.token_hex(4) # 8-char hex, e.g. "a3f1b2c8"
email = f"alice-{suffix}@example.com"
Use when: You want something shorter than a full UUID but still practically unique.
from faker import Faker
fake = Faker()
email = fake.unique.email()
name = fake.unique.name()
Use when: Tests need realistic-looking data, or you're already using Faker. The .unique property ensures no collisions within the generator session.
import factory
class UserFactory(factory.django.DjangoModelFactory):
class Meta:
model = User
email = factory.Sequence(lambda n: f"user{n}@example.com")
username = factory.Sequence(lambda n: f"user{n}")
# Usage — each call gets a unique email/username
user1 = UserFactory()
user2 = UserFactory()
Use when: You have many fields and many tests. factory_boy Sequences are the gold standard — zero boilerplate, fully automatic, works with SubFactory.
import pytest
@pytest.fixture
def unique_user():
suffix = secrets.token_hex(4)
user = User.objects.create(
email=f"test-{suffix}@example.com",
username=f"test-{suffix}",
)
yield user
user.delete()
Use when: Many tests need a user and you want DRY setup with guaranteed cleanup.
| ❌ Anti-Pattern | Why It Fails | ✅ Fix |
|---|---|---|
email="alice@example.com" | Second test or re-run collision | f"alice-{uuid.uuid4()}@example.com" |
User.objects.create() without cleanup in non-transactional tests | Data leaks between tests | Use @pytest.mark.django_db(transaction=True) or fixture with teardown |
User.objects.all().delete() at start of test | Race condition in parallel tests | Use secrets.token_hex() or factory |
| Copy-pasting seed data across test files | Same hardcoded values in multiple files | Extract a shared factory or fixture |
Using datetime.now() at second precision | Collision when tests run in same second | Use microsecond precision (%f) or UUID |
UserFactory() without unique guarantees on special fields | Factory inherits hardcoded defaults | Override with Sequence or pass unique values |
Before writing seed data, identify the model's unique constraints:
# Django
class Meta:
unique_together = [["organization", "slug"]]
constraints = [
models.UniqueConstraint(fields=["tenant", "external_id"], name="unique_tenant_external")
]
# SQLAlchemy
__table_args__ = (
UniqueConstraint("tenant_id", "external_id", name="uq_tenant_external"),
)
# Prisma / Rails / etc. — check schema.rb, schema.prisma, or migrations
Fields to always make unique:
id / pk (unless using auto-increment — then let the DB handle it)emailusernameslugexternal_id / stripe_id / github_id — any third-party IDapi_key / token / secretinvite_code / referral_codephone_numberhandle / display_handle / profile_urlunique=True, unique_together, or UniqueConstraintWhen the RED phase fails not because the test logic is wrong but because of a unique constraint violation:
When GREEN phase passes but seed data has hardcoded unique fields: this is a REFACTOR concern — dynamize the seed data while keeping tests green.
Before declaring seed data done:
✅ Every uniquely-constrained field uses a dynamic pattern (UUID, timestamp, counter, or Faker)
✅ No hardcoded alice@example.com, testuser, demo-slug, or similar in seed data
✅ Tests pass on re-run without cleanup
✅ Tests pass in parallel (xdist or similar)
✅ factory_boy fixtures use Sequence() for all unique fields
✅ Non-unique fields are still readable/human-friendly (don't over-randomize)