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.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
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.
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.
The Rule
Every value inserted into a uniquely-constrained field MUST be dynamically unique at runtime.
id / pk (unless using auto-increment — then let the DB handle it)
email
username
slug
external_id / stripe_id / github_id — any third-party ID
api_key / token / secret
invite_code / referral_code
phone_number
handle / display_handle / profile_url
Any field with unique=True, unique_together, or UniqueConstraint
Integration with change-test-loop
When the RED phase fails not because the test logic is wrong but because of a unique constraint violation:
Don't change the test logic — the test is correct
Fix the seed data to use a dynamic uniqueness pattern
Re-score RED: the test should now fail for the intended reason (feature missing)
When GREEN phase passes but seed data has hardcoded unique fields: this is a REFACTOR concern — dynamize the seed data while keeping tests green.
Verification
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)
