On November 10 the Chrome team released Modern Web Guidance (MWG) 1.0, the stable version of its skills pack for coding agents. It holds about 200 guides reviewed by browser engineers and accessibility experts, covering performance, accessibility, user experience, and security, with Baseline compatibility data built in, so agents write less outdated JavaScript and use native HTML and CSS features instead. Version 1.0 adds a modernization workflow that statically scans existing HTML, CSS, JavaScript, and package.json for legacy patterns that can be safely upgraded. The project is also no longer Chrome-only: Firefox and Edge now contribute. It works with Antigravity, Cursor, GitHub Copilot, Codex, WebStorm, and more, and installs with one command: npx modern-web-guidance@latest install.
Key Takeaways
- ✓Scale: about 200 expert-reviewed guides, roughly 20% of which
legacy-scancan check deterministically; from 1.0 every guide has evals, with a public eval dashboard (official figures) - ✓Scan rules: tuned and tested on about 45,000 open-source repos split in half; default rules must hit 80%+ static matcher precision and 90%+ post-verification precision; local AST parsing with zero LLM calls
- ✓Case study: Tencent WeTV added it to CI/CD and its agent now resolves 90% of failed performance checks, cutting fix time from 4 hours to 30 minutes (87% less, vendor figures)
- ✓Case study: across 180 eval runs on its production codebase, note inc. saw modern web check pass rates rise from 61% to 94%, and LCP image prioritization improved 28% across 431 merged PRs
- ✓Getting started:
npx modern-web-guidance@latest install, orupdatefor older installs; type/modern-web-guidance Modernize the code in this directoryin your agent to start the modernization workflow
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background and Industry Pain Points
Frontier coding models can write frontend code that works, but much of their training data reflects patterns from years ago: dialogs, tooltips, and drawers often pull in heavy libraries or hundreds of lines of custom JavaScript, even though browsers now support dialog, popover, and scroll-driven animations natively. Chrome's evals found models often get keyboard focus management, reduced motion, CSP, HTML sanitization, and passkeys wrong, and bad ARIA can be worse than none.
Architecture Highlights and Underlying Mechanisms
MWG plugs into agents as skills that retrieve guides on demand. Each guide embeds Baseline and browser compatibility data, so advice matches the project's support targets. The new modernization workflow in 1.0 first scans code locally with AST, DOM, and CSS parsers at no token cost; each match carries keep and drop rubrics so the agent can filter false positives in context before guiding the changes. The project moves from Chrome-only to a joint effort with Firefox and Edge.
Authoritative Benchmarks and Measured Comparisons
Every guide now has evals, with a public dashboard comparing agents with and without MWG. The following are vendor case-study figures: note inc. raised modern web check pass rates from 61% to 94%; Tencent WeTV auto-resolves 90% of failed performance checks and cut fix time by 87%; Pensionbazaar shortened new-page turnaround by more than 90%. No independent replication has been published yet.
Developer Adoption and Getting Started
Run npx modern-web-guidance@latest install in your project directory; the CLI route enables automatic updates, and agent plugin marketplaces also carry it. For existing projects, run legacy-scan directly or ask your agent to modernize the directory. Chrome recommends pairing it with Chrome DevTools for agents to verify changes with performance traces and screen recordings, for example keeping main-thread tasks under 50 milliseconds.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.