Test Data Cleanup Skill

Golden Rule

All test data MUST contain "Test" in the name or "@test." in email.

This ensures safe, pattern-based cleanup that never touches real data.

Tenant Structure (CRITICAL)

Production Tenant

• Volta (11111111-1111-1111-1111-111111111111) - Main org with REAL data mixed with test data
• Contains: Real companies, contacts, Fireflies meetings, AND test forms/programs

Test-Created Tenants (Empty Shells)

Tests often create new tenants instead of using Volta. These are identifiable by:

• Timestamp in name: Org A 1763320623066
• Generic test names: Test, Organization A, Organization B
• All have 0 records - safe to delete entirely

Predefined Test Tenants (May Not Exist)

• Acme/Volta: 11111111-1111-1111-1111-111111111111
• Beta: 22222222-2222-2222-2222-222222222222
• Gamma: 33333333-3333-3333-3333-333333333333

Cleanup Strategy

Step 1: Analyze Tenant Structure

-- Show all tenants with data counts
SELECT
  t.name as tenant_name,
  t.id as tenant_id,
  CASE
    WHEN t.id = '11111111-1111-1111-1111-111111111111' THEN 'VOLTA (Main)'
    WHEN t.name ~ '\d{10,}' THEN 'TEST-CREATED (timestamp)'
    WHEN t.name IN ('Test', 'Organization A', 'Organization B') THEN 'TEST-CREATED (generic)'
    ELSE 'UNCLEAR'
  END as tenant_type,
  (SELECT COUNT(*) FROM companies c WHERE c.tenant_id = t.id) as companies,
  (SELECT COUNT(*) FROM contacts co WHERE co.tenant_id = t.id) as contacts,
  (SELECT COUNT(*) FROM forms f WHERE f.tenant_id = t.id) as forms
FROM tenants t
ORDER BY tenant_type, t.name;

Step 2: Delete Empty Test-Created Tenants

-- Find empty test-created tenants
SELECT id, name FROM tenants
WHERE (name ~ '\d{10,}' OR name IN ('Test', 'Organization A', 'Organization B'))
  AND id != '11111111-1111-1111-1111-111111111111'
  AND id != '22222222-2222-2222-2222-222222222222'
  AND id != '33333333-3333-3333-3333-333333333333';

-- Delete them (after verification)
DELETE FROM tenants WHERE id IN (...);

Step 3: Clean Test Data in Volta (Pattern-Based)

Only delete records with "Test" in name/title:

-- Preview counts in Volta
SELECT 'forms' as table_name, COUNT(*) FILTER (WHERE title ILIKE '%test%') as test_matches
FROM forms WHERE tenant_id = '11111111-1111-1111-1111-111111111111'
UNION ALL
SELECT 'programs', COUNT(*) FILTER (WHERE name ILIKE '%test%')
FROM programs WHERE tenant_id = '11111111-1111-1111-1111-111111111111'
UNION ALL
SELECT 'report_sessions', COUNT(*) FILTER (WHERE title ILIKE '%test%')
FROM report_sessions WHERE tenant_id = '11111111-1111-1111-1111-111111111111'
UNION ALL
SELECT 'companies', COUNT(*) FILTER (WHERE business_name ILIKE '%test%')
FROM companies WHERE tenant_id = '11111111-1111-1111-1111-111111111111';

Step 4: Delete Test Users

-- Find test users by email pattern
SELECT id, email FROM auth.users
WHERE email LIKE '%@test.%' OR email LIKE '%@example.%';

Tables to Clean (FK-Safe Order)

Order	Table	Pattern Column	Query
1	forms	title	title ILIKE '%test%'
2	report_sessions	title	title ILIKE '%test%'
3	programs	name	name ILIKE '%test%'
4	companies	business_name	business_name ILIKE '%test%'
5	contacts	first_name/last_name	first_name ILIKE '%test%' OR last_name ILIKE '%test%'
6	user_sessions	(linked)	Delete for test users
7	organization_members	(linked)	Delete for test users
8	auth.users	email	email LIKE '%@test.%'
9	tenants	name	Empty test-created tenants

Tables to NEVER Clean (Real Data)

Table	Reason
fireflies_staged_meetings	Real imported meetings
meeting_transcripts	Real meeting data
interactions	Real meeting records
contacts (without Test)	Real people
companies (without Test)	Real businesses
reports	Real generated reports

Naming Conventions (For Writing Tests)

Users/Emails

// CORRECT - contains @test.com
`test-feature-${Date.now()}@test.com`

// WRONG - real domain
`user-${Date.now()}@gmail.com`

Companies

// CORRECT - contains "Test"
`Test Company ${Date.now()}`

// WRONG - no "Test" identifier
`Company A ${Date.now()}`

Contacts

// CORRECT - contains "Test"
{ first_name: 'Test', last_name: `User ${Date.now()}` }

// WRONG - looks like real person
{ first_name: 'Alice', last_name: 'Smith' }

Tenants

// CORRECT - use existing test tenant
const tenantId = TEST_TENANT_ACME // '11111111-1111-1111-1111-111111111111'

// WRONG - creating new tenant
const { data: tenant } = await adminClient.from('tenants').insert({ name: 'Org A' })

Agent Workflow

When asked to clean up test data:

1. Analyze - Query tenant structure and data distribution
2. Identify - Find test-created tenants and test pattern records
3. Preview - Show summary table to user
4. Confirm - Ask for explicit approval
5. Delete - Execute in FK-safe order
6. Report - Show results

Example Output

## Tenant Analysis
- Volta (main): 283 companies, 836 contacts, 318 Fireflies meetings
- Test-created tenants: 13 (all empty)

## Test Data in Volta
- Forms with "Test": 203
- Programs with "Test": 10
- Report sessions with "Test": 18
- Companies with "Test": 3

## Test Users
- @test.com/@example.com emails: 527

## Recommended Cleanup
1. Delete 13 empty test-created tenants
2. Delete 527 test users
3. Delete 203 test forms
4. Delete 10 test programs
5. Delete 18 test report sessions
6. Delete 3 test companies

Proceed? (y/n)

Test Helper Integration

Use __tests__/api/helpers.ts:

import {
  createTestUser,           // Uses @test.com email
  createTestCompany,        // Defaults to "Test Company..."
  createTestContact,        // Defaults to "Test" first name
  createTestForm,           // Defaults to "Test Form..."
  createTestReport,         // Defaults to "Test Report..."
  countTestDataByPattern,   // Preview before delete
  cleanupTestDataByPattern, // Safe pattern-based cleanup
  TEST_TENANT_ACME,         // Use this, don't create new tenants!
  TEST_TENANT_BETA,
  TEST_TENANT_GAMMA
} from '@/__tests__/api/helpers'

Common Issues

Issue: Tests create new tenants

Fix: Always use TEST_TENANT_ACME instead of creating new tenants

Issue: Test data doesn't have "Test" in name

Fix: Update test to use proper naming convention

Issue: Real data accidentally has "Test" in name

Check: Verify before deletion - this should be rare