Build better websites with the Web Development AI Skills Toolkit! Review accessibility, security, performance, technical SEO, launch readiness, and overall project health, with readable reports, screenshots, and clear action steps.
Web Development AI Skills Toolkit helps developers review websites without starting from a blank page—or working through hundreds of tiny checklist items.
It combines human-readable checklists with automated evidence collection and AI-assisted review. Work through a checklist yourself, run a website report from the command line, or ask Codex to inspect the available evidence and recommend practical next steps.
The checklists remain the source of truth. Automation covers the checks it can actually perform, and the agent explains what still needs access, context, or a human decision.
Web developers, engineers, technical project leads, and small teams maintaining websites, inheriting projects, or preparing a release. General reviews work across frameworks and CMS platforms; the documentation also includes WordPress lifecycle checklists and Sitefinity maintenance guidance.
* Six Codex / local AI-agent review skills: Aaccessibility, security, performance, technical SEO, launch readiness, and a cross-discipline project audit.
* Human-runnable reports: Run the same deterministic evidence profiles from the command line, without an AI session.
* Readable evidence: HTML reports, JSON results, page screenshots, and bounded screenshots of elements flagged by axe.
* Shared browser tooling: Playwright, Chromium, axe, and Lighthouse, with dependencies cached outside the website project.
* Desktop audit baseline: Repeatable browser settings, with mobile, zoom, and interactive testing kept separate.
* Practical documentation: Development and launch checklists, WordPress setup/testing/migration guidance, and engineering workflows.
* Honest coverage: Passes, findings, warnings, and unverified checks are kept distinguishable. Launch recommendations include remaining human checks.
* Organized output: Timestamped evidence under `.output////` by default.
* Automated accessibility checks do not certify WCAG conformance or replace keyboard and assistive-technology testing.
* The Lighthouse reports run in desktop mode and do not establish mobile performance, real-user Core Web Vitals, or geographic performance.
* A public URL cannot verify backups, restore procedures, administrator access, private configuration, or business workflows. Those checks need additional evidence.
* Live content, network conditions, scores, and agent judgments can vary even though the collection procedure is repeatable.
Start with the repository and setup instructions. You can use the human checklists without installing anything.
For command-line reports, clone the repository and open a terminal in its root. Check runtime readiness and, on first use, approve and run the bootstrap:
npm run runtime:status
npm run runtime:bootstrap
Bootstrap downloads the pinned audit dependencies and matching Chromium into a user cache. Once ready, run a profile:
npm run review:website -- --profile review-web-accessibility --url https://site.example
Replace https://site.example with a website you are authorized to review. Open accessibility-report.html in the resulting .output directory. Other profiles create their own report files.
For Codex, use the desktop app’s Plugins → Add plugin marketplace dialog to add compkick/web-development-ai-skills-toolkit directly from GitHub. Select Personal and install Web Development AI Skills Toolkit, then start a new chat. No separate CLI installation or manual clone is required for this route.
Try a request such as:
Review this website’s accessibility, save the evidence under .output, and give me the most important fixes first.
To invoke a specific skill, for example, use $review-web-accessibility in your request. The other five skill names and CLI profiles are listed in the README.
First-use approvals may cover downloads, browser execution, network access, and evidence files. Choose the permission mode you understand and are comfortable with; full access is not required.
Review reports before sharing them, because screenshots and URLs may contain sensitive information.
* Checklists only: a browser or Markdown reader. No Codex, Node.js, or Python installation is needed.
* Automated reports: Node.js 22.19.0 or newer, npm, an approved Chromium download or a compatible installed Chrome/Edge browser, and network access to the authorized target.
* AI-assisted reviews: a working Codex environment with plugin support and the toolkit installed. Codex usage is separate from the toolkit’s MIT license.
* Platform status: Windows 11 x64 and Ubuntu CI are tested, including all six browser-profile suites. macOS, Claude Code, and Bionic compatibility are unverified. See the supported platform matrix for exact coverage and remaining release tests.
* Contributors: repository validation dependencies are installed with npm ci. Python and PyYAML are for the official skill/plugin validators, not ordinary website reports.
This project is licensed under the MIT License. You are free to use, modify, and distribute the software under the terms of the included license.
The software is provided “as is”, without warranty of any kind, and the authors are not liable for any claims, damages, or other liability arising from its use.
See project documentation in the repo README.md, which covers installation, setup and usage.
For implementation questions, issue reporting, or enhancement requests, use the GitHub issue tracker:
https://github.com/compkick/mcp-sql-tiny
For other questions, use our contact form to get in touch.