clone-testing

name	clone-testing
description	Comprehensive testing -- unit, E2E, visual comparison with original, performance, security

Clone Testing

When to Use

• MVP implemented (output of clone-implementation)
• Need to verify correctness, performance, and visual fidelity
• Pre-deployment quality gate

Process

1. Unit tests -- core business logic:
   • Model validation and transformations
   • Utility functions
   • AI prompt construction
   • Target: 80%+ coverage on core modules
2. Integration tests -- API layer:
   • Each endpoint: happy path + error cases
   • Auth flow end-to-end
   • Database state verification
   • AI provider mock for deterministic results
3. E2E tests (Playwright) -- user flows:
   • Critical path: registration --> create --> edit --> export
   • Navigation and routing
   • Form submissions and validation messages
   • Responsive behavior (desktop + mobile viewports)
4. Visual comparison -- clone vs original:
   • Screenshot same pages on original and clone
   • Side-by-side comparison (manual or pixel-diff tool)
   • Document deviations with rationale (intentional vs bug)
5. Performance benchmark:
   • Lighthouse score (Performance, Accessibility, SEO)
   • API response times (p50, p95, p99)
   • AI generation latency (time-to-first-token, total)
   • Bundle size analysis
6. Security scan:
   • Dependency audit (npm audit / pip-audit)
   • OWASP Top 10 checklist
   • Auth bypass attempts
   • Input injection tests (XSS, SQL injection, prompt injection)

Output

• Test suite (runnable via pytest / npx playwright test)
• Visual comparison report (screenshots + diff annotations)
• Performance metrics table
• Security audit report with severity ratings
• Overall quality verdict: PASS / PASS WITH NOTES / FAIL

Example

SlideDeck AI testing: 47 unit tests (92% coverage), 12 integration tests, 6 E2E flows. Visual comparison shows 85% fidelity to Gamma.app -- deviations in animation timing (intentional) and font rendering (browser difference). Lighthouse: 89 Performance, 95 Accessibility. Security: 0 critical, 2 medium (rate limiting missing on AI endpoint, CSP header incomplete).