| name | quality-and-verification |
| description | 5-level verification pyramid: static→unit→Playwright E2E (homepage-first, 6bp)→AI visual→post-deploy. 8-check quality gate. Multi-agent testing (functional/security/a11y/performance). Playwright v1.59+ AI agents (Planner/Generator/Healer). WCAG 2.2 AA via axe-core v4.11. Percy+Chromatic visual regression. ADA Title II 2027/2028 deadlines. |
| metadata | {"version":"2.0.0","updated":"2026-04-23","token_budget":"5K"} |
| license | Rutgers |
| compatibility | {"claude-code":">=2.0.0","agentskills":">=1.0.0"} |
| submodules | ["accessibility-gate.md","performance-optimization.md","security-hardening.md","computer-use-automation.md","chrome-and-browser-workflows.md","completeness-verification.md","visual-inspection-loop.md","tdd-verification.md","testing-matrices.md","adversarial-testing.md","spec-driven-development.md","stagehand-ai-testing.md","visual-regression.md","contract-testing.md","slop-detection.md","eval-driven-development.md","ui-completeness-sweep.md","semgrep-codebase-rules.md","audio-video-sync.md","e2e-accumulation.md","evidence-collection.md","agentic-security.md","picovoice-eagle-biometric.md","stagehand-ai-fallback.md"] |
07 — Quality and Verification
Submodules: accessibility-gate (axe-core v4.11, WCAG 2.2 AA, focus-not-obscured, target-size 24px, accessible-auth)|performance-optimization (CWV, INP 3-phase, budgets)|security-hardening (CSP nonce-based, OWASP 2025, supply-chain #3)|computer-use-automation (native macOS)|chrome-and-browser-workflows (Chrome/Playwright MCP)|completeness-verification (multi-pass AI visual)|visual-inspection-loop (screenshot/critique/fix)|tdd-verification (10-point journey)|testing-matrices (payment/email/form)|adversarial-testing (chaos/stress)|stagehand-ai-testing (AI browser fallback, a11y-tree selectors)|visual-regression (Percy AI+Chromatic+pixelmatch)|contract-testing (Zod vs live APIs)|slop-detection (AI filler scanner)|eval-driven-development (LLM-as-judge)|ui-completeness-sweep (MANDATORY Playwright+GPT-4o, blocks done until >=8/10)|semgrep-codebase-rules (AST-level per-project rules)|e2e-accumulation (append-only parallel chunks)|evidence-collection (R2 video+screenshots).
Verification Pyramid
L5: Post-Deploy (08). L4: Visual (screenshots+AI). L3: E2E (Playwright v1.59+). L2: Integration (API/DB). L1: Static (lint/typecheck).
Every code change: L1-L3. Every deploy: L4-L5.
Playwright AI Agents (v1.59.1+)
Planner: explores app, designs test plans from natural language. Generator: creates executable test code. Healer: auto-fixes broken tests. Pattern: static specs for stable tests, AI agents for flaky/new. Run agents only on failed tests in second pass (70% token savings).
Playwright v1.59.1 New APIs
page.screencast({ path }) — video recording with action annotations and real-time frame capture (video receipts for CI). browser.bind() — connect to running browser instances. page.consoleMessages()/page.pageErrors()/page.requests() — snapshot-in-time accessors (no event listeners needed). await using async disposables for auto-cleanup. Trace CLI for agent-driven test analysis.
MCP-Based Testing
Playwright MCP operates on accessibility tree, not screenshots. Returns structured snapshots: role hierarchy, names, states. Target "Role: button, Name: Checkout" — 10x more stable than CSS selectors. AOM-reasoning > DOM-scraping.
Multi-Agent Testing Pattern
Functional agent: happy path clicks. Security agent: XSS probing, auth bypass. Accessibility agent: WCAG 2.2 compliance. Performance agent: CWV measurement. Run all four in parallel per deploy.
Self-Healing Loop
Test fails→read full error→classify (code/test/env/flaky)→fix root cause→re-run→loop max 5. NEVER .skip/.only/.fixme()/comment assertions/increase timeouts. After fix: full suite.
Zero Console Errors
: [] = [];
page.(, { (msg.() === ) errors.(msg.()); });
(errors).([]);
