Website Benchmark — Security skill for Claude Code

website-benchmark

A [Claude Code](https://claude.com/claude-code) skill that turns **one website URL** into:

1. Research brief: audience, ranked visitor needs, a market profile for each country (population mix, languages, devices, social platforms, holidays, climate, payments, laws, maps, local competitors, search behaviour), a 12-month seasonal calendar, best practices and trends. Every point is cited.
2. Benchmark report: the site scored against 8–20+ best-in-class competitors on 8 dimensions. Each issue is laid out as Problem → Screenshot → Who does it better → Research reason → Fix, followed by a roadmap.
3. Component kit: ready-made UI components (HTML, React, copy-paste CSS) for each issue. Each one has a screenshot, the exact location on its source page, the licence and a real-world reference.
4. CSS / UI resource directory: vetted component sites with verified copy buttons for GitHub repos, agent skills, MCP servers, npm packages and llms.txt files.
5. Build prompt: a ready-to-paste prompt for another Claude session to build the new site from all of the above.

How it works

1. Intake: you give a URL plus any extras (crawl folder, analytics exports, brand guidelines, notes).
2. Questions: multiple choice and pre-filled from the site: website type, goal, visitors, countries, competitor markets, languages, key pages, research and benchmark depth, outputs, browsing permissions.
3. Scope check, twice: a full proposal you can edit, then Start · Edit again · Cancel. Nothing heavy runs before you choose Start.
4. Research: parallel agents cover geography and local market, industry and audience, beyond-the-website signals, and competitor discovery.
5. Audit: a technical script, desktop and mobile screenshots in every language, a visual review and the research checklist.
6. Benchmark: parallel agents by region, all using the same template.
7. Scoring: an 8-dimension rubric weighted by the research.
8. **Outpu