Modern Web Guidance 1.0: The web platform expertise your coding agent is missing | Blog | Chrome for Developers

Open original

Chrome, Firefox, and Edge are backing Modern Web Guidance 1.0, a set of skills that feed web platform expertise, best practices, and browser compatibility data into AI coding agents. The release adds a codebase modernization workflow and pairs with Chrome DevTools for agents so the agent can verify its own output.

  • Coding agents often miss accessibility, performance, security, and UX details, such as focus management, main-thread blocking over 50ms, and CSP setup, so the guidance targets those gaps.
  • The guides steer agents toward native HTML and CSS features like popovers, dialogs, and autofill over custom JavaScript, based on the browser support a project needs.
  • Version 1.0 adds a static-analysis modernization workflow that scans HTML, CSS, JavaScript, and package.json for outdated patterns and polyfills, though it covers only a limited set of patterns for now.

Modern Web Guidance is a set of skills that feed web platform expertise, best practices, and browser compatibility data into coding agents, and version 1.0 went stable with three main additions. There's a modernization workflow that scans existing HTML, CSS, JavaScript, and package.json files for outdated workarounds and unneeded polyfills, though in this first release it relies on static analysis and only checks a limited set of patterns. The guide database has expanded to hundreds of verified knowledge gaps, each reviewed by content-area tech leads, with a dedicated accessibility review on user-facing guides. And Firefox and Edge have joined Chrome as contributors, which matters because the guidance is meant to work regardless of which model or agent harness a team uses. Install is a single `npx modern-web-guidance@latest install` command, with `update` for existing setups. The argument behind the release is that AI coding tools are fast but often miss things an experienced reviewer would catch. Chrome's evals flag specific gaps: keyboard focus and inert management, reduced-motion support, and accessible form validation on the accessibility side; main-thread task scheduling and layout shift prevention on performance; and HTML sanitization, strict Content Security Policies, and passkey handling on security. The authors also make a point about accessibility: bad ARIA can leave a page worse off than no ARIA at all, which is why they had real assistive technology users test the user-facing guides. A related theme is preferring native platform features over hand-rolled JavaScript. Popovers, dialogs, scroll-driven interactions, and view transitions often need far less code, run better, and come with accessibility built in. Autofill and passkeys are highlighted for checkout and login flows, where custom JavaScript form controls tend to break. The most practical part is the pairing with Chrome DevTools for agents. Guidance tells the agent how to write something, while DevTools lets it measure whether the result works, such as running a performance trace on a simulated low-end CPU to see how long tasks really block the main thread. The write-then-verify loop is the point, and the authors argue it gets to a better result in fewer turns. For anyone already using coding agents, the cost of trying it is one command, and the modernization workflow is the piece most worth running against an existing codebase to see what it finds.

Ask about this article