Coding agents write code fast, but not always well. On Tuesday Chrome shipped the stable 1.0 release of Modern Web Guidance, a set of skills that embed web platform expertise, best practices and browser compatibility data directly into AI coding agents - and it is now a cross-browser project, with Firefox and Edge joining Chrome.

Modern Web Guidance began as an early preview at Google I/O in May. The 1.0 release adds four things: a modernisation workflow that scans existing HTML, CSS, JavaScript and package.json for opportunities to safely upgrade; an expanded guide database targeting hundreds of verified AI knowledge gaps; deeper human expert review, including sign-off from content-area tech leads and a dedicated accessibility review on user-facing guides; and full evaluation coverage with a public dashboard showing measurable improvements over agents without the guidance.

The pitch is that frontier models have real blind spots. Google's evals flag accessibility (keyboard focus and inert management, reduced-motion support, accessible form validation), user experience (swipe and touch gestures, dialog and popover transitions, light/dark themes), performance (main-thread task scheduling, offscreen rendering isolation, layout shift prevention) and security (safe HTML sanitization, strict Content Security Policies, passkey credential management). The tool also nudges agents toward native HTML and CSS instead of JavaScript for the same job.

Installation is a single command - npx modern-web-guidance@latest install - and the skills also ship through the plugin marketplaces of Claude Code, Codex, Cursor, Antigravity, Copilot, WebStorm, Kimi Code and Grok Build. Chrome positions it alongside Chrome DevTools for agents: guidance to write the code, DevTools to run and verify it.

Google cites production numbers from early adopters. note inc., a Japanese creator platform with more than 90 million monthly active users, says a 180-run evaluation on its production codebase raised pass rates on modern web checks from 61% to 94% and increased proper LCP image prioritization by 28% across 431 merged pull requests. MetaMask runs the guidance as an automated review bot on every pull request; in its first five weeks the bot reviewed about 480 PRs, commented on 64 and led engineers to fix 43 findings with a median time of 15 minutes. WeTV, Tencent's international streaming platform, says an agent using the tooling now auto-resolves 90% of failed performance checks. These figures come from Google and its partners.

In a world where everyone uses coding agents, Google argues, speed is no longer the differentiator - quality of experience is. The open-source source repository, GoogleChrome/modern-web-guidance-src, was among GitHub's fastest-moving repos this week.