Cookbook

QA Audit — Live Site

Run the 7-dimension QA orchestrator against a live deployed site

QA Audit — Live Site

This recipe walks through a full 7-dimension QA audit using the QA orchestrator (marketing-website-qa plugin) against a live deployed site: https://registry.vandoko.com.

Prerequisites: Claude Code with the marketing-website-qa plugin active. A site directory or live URL to audit. Optionally: a globals.css path for brand consistency audit, and a voice-spec.md for the content quality gate.


Command

Terminal
/qa audit site

Or: "run QA", "full site audit", "audit the site", "is the site ready"


Pipeline

Step 1: Gather Input

The QA orchestrator asks for 4 fields:

FieldYour Test Answer
Site directory~/dev/projects/vandoko/vandoko-registry (or provide deployed URL)
SiteBlueprintno (skip)
Voice specno (skip)
Design tokens~/dev/projects/vandoko/vandoko-registry/src/app/globals.css

Step 2: Determine Scope

System asks: "What type of audit?" Answer: Full site audit

System confirms: "Running all 7 audit skills. Proceed?" Answer: yes

Step 3: Route to Skills

Each skill runs sequentially and produces an audit_result:

#SkillKeyWhat It Checks
1Visual Quality Gatevisual_qualityLayout consistency, spacing rhythm, color harmony, responsive breakpoints, visual hierarchy
2Brand Consistency Auditbrand_consistencyoklch token usage, design system compliance, color token drift, font stack adherence
3SEO AuditseoMeta tags, structured data, sitemap, canonical URLs, heading hierarchy, image alt text
4Accessibility AuditaccessibilityWCAG 2.1 AA compliance, color contrast, keyboard navigation, ARIA labels, focus management
5Conversion AuditconversionCTA visibility, form usability, trust signals, friction points, above-fold clarity
6Content Quality Gatecontent_qualityCopy clarity, brand voice adherence, readability score, content completeness
7Production Readinessproduction_readinessBuild errors, console warnings, broken links, asset optimization, security headers

Step 4: Aggregate into QualityReport

The 7 dimension scores are combined into a single QualityReport. production_readiness is pass/fail and is not included in the weighted score calculation.

How to interpret scores:

Score RangeGradeMeaning
90–100ALaunch-ready, minimal polish needed
80–89BSolid, minor improvements recommended
70–79CFunctional but needs work before launch
60–69DSignificant issues, do not launch
Below 60FMajor problems, requires substantial rework

Step 5: Recommendation

Based on the score, the system determines the next action:

  • overall_score >= 80 AND production_readiness.passed = trueLaunch clearance. System says: "Site is cleared for launch."
  • overall_score >= 80 AND production_readiness NOT yet run → System recommends: "Run production-readiness for pre-launch sign-off."
  • overall_score < 80Polish pass triggered. System recommends: "Score below launch threshold. Run site-polish for iterative fix passes."

Verification

  • QualityReport JSON has all 7 dimension scores
  • overall_score = weighted sum (visual 25%, SEO 20%, accessibility 20%, brand 15%, conversion 10%, content 10%)
  • production_readiness is pass/fail (not included in weighted score)
  • top_issues has up to 10 items, sorted: critical > major > minor
  • Each issue has location, description, and fix fields
  • Grade matches score range table

Troubleshooting

SymptomLikely CauseFix
brand_consistency score is 0 or skippedNo design tokens path providedProvide globals.css path; weight redistributed if skipped
content_quality not runningNo voice spec providedOptional; skill runs with generic voice rules when no spec is given
production_readiness blocks on build errorsNext.js build failsFix build errors first; re-run production-readiness alone
Overall score seems wrongWeight redistributionIf a dimension is skipped, its weight redistributes evenly across remaining dimensions

On this page