Phase 7 — Automated Quality Loop

This is a LOOP. Do NOT exit until ALL automated conditions are satisfied. Visual analysis is handled by Phase 7b (separate agent with fresh context).

Objective

Validate the built site across automated scoring, blueprint adherence, performance, and SEO — looping until every automated gate passes. This phase focuses on TECHNICAL quality. Visual quality and enchantment are evaluated in Phase 7b.

Pre-Gate

.claude/scripts/gate-quality-loop.sh $LEAD_ID --pre

If the pre-gate fails (mapa-encantamento.md missing), STOP. Return to Phase 3.

Gate 1 — Automated Score >= 9.0

Start dev server in background:

cd /Users/felipemoreiralanna/Documents/GitHub/vendedor-de-sites-v2/sites/$LEAD_ID && npm run dev

(Use run_in_background: true. Wait 5 seconds.)

Run validation (31 automated UX/UI checks, score 0-10):

cd /Users/felipemoreiralanna/Documents/GitHub/vendedor-de-sites-v2
.venv/bin/python sites/_templates/playwright-validate.py http://localhost:5180 sites/$LEAD_ID/screenshots

If score < 9.0 (exit code 1):

1. Read EVERY red-CRITICAL and FAIL item
2. Fix in code
3. Kill server: pkill -f "vite.*5180" 2>/dev/null; true
4. Restart: cd sites/$LEAD_ID && npm run dev (background)
5. Re-run script
6. LOOP until score >= 9.0

Gate 1.1 — 34-Criteria System

Load the complete criteria reference:

Read: sites/_templates/visual-review-criteria.md

The playwright-validate.py script scores 31 automated criteria. You must also evaluate criteria 32-34 manually:

Load rubrics:

Read: sites/_templates/quality-rubric.md

Apply criteria 32 (Visual Originality), 33 (Desktop Professionalism), 34 (Mobile Professionalism).

Approval rules:

• EVERY one of the 34 criteria >= 8
• Average of all 34 >= 9.0
• ANY originality criterion (29, 30, 32) < 8 → FAIL
• ANY critical criterion = 0 → STOP immediately

Gate 1.2 — Blueprint Adherence

Read: sites/$LEAD_ID/mapa-encantamento.md

For EACH section in the blueprint, verify:

1. Was the specified LAYOUT implemented in code?
2. Does the specified ANIMATION exist in the component?
3. Was the VISUAL TECHNIQUE applied?

If ANY section diverges from the blueprint → FIX and re-validate.

Gate 1.5 — Performance Profiling (optional)

Only execute if Chrome DevTools MCP is connected (Chrome running with chrome-debug).

Metric	Good	Acceptable	Action if exceeded
LCP	< 2.5s	< 4.0s	Optimize blocking assets
CLS	< 0.1	< 0.25	Fix layout shifts
INP	< 200ms	-	Reduce JS blocking

If Chrome DevTools NOT available: skip and log "Gate 1.5 skipped — Chrome DevTools not available."

Gate 4 — SEO Audit

After Gates 1-1.5 pass, invoke: /seo page http://localhost:5180

For EACH issue found:

1. Fix in code (meta tags, schema, headings, alt text, hreflang, sitemap, etc.)
2. Re-run /seo page http://localhost:5180 to confirm
3. LOOP until SEO Health Score has zero criticals

After SEO fixes, re-run playwright-validate.py to confirm score >= 9.0 STILL holds. Re-validate mobile.

Server Cleanup

Kill dev server when exiting: pkill -f "vite.*5180" 2>/dev/null; true

NOTE: Do NOT kill the server if Phase 7b will run immediately after — leave it running for the visual review agent.

Constraints

Constraint	Enforced by
Score >= 9.0	playwright-validate.py exit code
All 34 criteria >= 8	Criteria check + visual-review-criteria.md
Blueprint adherence	Manual section-by-section check
SEO no criticals	/seo page audit
Mobile parity	Every fix requires mobile re-validation

Exit Criteria

ALL must be TRUE:

• Gate 1: Automated score >= 9.0, zero CRITICAL checks
• Gate 1.1: All 34 criteria >= 8, average >= 9.0
• Gate 1.2: Every blueprint section faithfully implemented
• Gate 1.5: CWV within limits OR "skipped — Chrome DevTools not available"
• Gate 4: SEO audit with zero criticals, fixes re-validated
• Dev server left running for Phase 7b (or killed if orchestrator handles restart)