# website-evaluator A Docker-based website audit tool. Given a URL it captures screenshots, runs Lighthouse audits (desktop + mobile), extracts DOM metadata, and assembles a structured Markdown report designed to be read by Claude for design critique and improvement recommendations. ## Quick Start ```bash # Build the image docker build -t website-evaluator . # Evaluate a site mkdir -p output docker run --rm -v "$(pwd)/output:/output" website-evaluator https://example.com # Feed the report to Claude # Open output/report.md and paste it into a Claude conversation with: # "Here is a website evaluation report. Please analyse it and suggest improvements." ``` ## Output ``` output/ report.md # Primary artifact — structured Markdown for Claude screenshots/ desktop.png # Full-page desktop (1440×900) mobile.png # Full-page mobile (375×812) raw/ lighthouse-desktop.json lighthouse-mobile.json dom.json ``` ## What the Report Covers - **Lighthouse scores** — Performance, Accessibility, Best Practices, SEO (desktop + mobile) - **Core Web Vitals** — LCP, TBT, CLS, TTFB with status labels - **Critical issues** — Pre-interpreted findings with "so what" explanations and fix guidance - **Page structure** — Title, meta description, canonical, heading hierarchy, images, links - **Accessibility detail** — WCAG violations grouped by impact level - **Performance detail** — Opportunities and diagnostics with estimated savings - **SEO detail** — Meta, robots, canonical, structured data ## Milestones | # | Scope | Status | |---|---|---| | M1 | Core pipeline: screenshot + Lighthouse + DOM → report.md | ✅ Complete | | M2 | `--analyze` flag: Claude API integration, AI recommendations | Planned | | M3 | Comparison mode: before/after and competitor diffs | Planned | ## Scripts | Script | Purpose | |---|---| | `scripts/verify-m1.sh` | Smoke test: builds image, runs against example.com, checks all outputs exist | ## Architecture ``` URL in └─▶ Playwright: screenshot desktop (1440px) + mobile (375px) + DOM extraction └─▶ Lighthouse: desktop audit └─▶ Lighthouse: mobile audit └─▶ Report assembly → /output/report.md ``` Built on `mcr.microsoft.com/playwright/node:20-noble` — Chrome is bundled, no separate install needed.