Image: GoogleChrome
Why it mattersA coding agent left on its own picks a stale pattern. A small rule pack it reads before writing changes what gets written.
A coding agent writes what it has seen most often. Most of what it has seen is old. Philip Walton's post on the Chrome for Developers blog announced Modern Web Guidance 1.0. It is a free rule pack for coding agents. The Chrome, Microsoft Edge, Mozilla Firefox and MDN teams look after it together with the web development community.
A team installs it with npx modern-web-guidance@latest install. It plugs into the agents a team already uses. The announcement names Claude Code, Codex, Cursor, Antigravity, Copilot, WebStorm, Kimi Code and Grok Build.
What it gives an agent
The pack is a set of rules the agent reads before it writes code. The GitHub repository says it covers 250 modern web platform features. It also lists 203 real-world developer use cases. It carries a legacy-scan command. The command reads an existing codebase. It reports where the code still uses older patterns. The checks cover accessibility, performance, user experience and security. The maintainers say each rule was reviewed by a domain lead.
Walton describes four kinds of work the pack helps with. The agent follows the rules while writing new code. It scans an existing codebase for legacy patterns. It checks a change for accessibility problems before the pull request lands. It reports on performance metrics the rules name, including Largest Contentful Paint and Interaction to Next Paint. The 1.0 release is the first stable version. A preview launched earlier in the year at Google I/O.
How some early users describe it
Walton's post names four teams. Pensionbazaar says the pack cut turnaround on front-end work by more than 90 percent. It says UI bugs fell by over 90 percent. WeTV says it closed 90 percent of failed performance checks on its own. WeTV says debugging time fell by 87 percent. The post says note inc. moved from 61 percent to 94 percent on its modern-web checks. MetaMask says its review bot processed about 480 pull requests. Its engineers fixed the issues the bot raised in a median of 15 minutes. These are the teams' own figures.
Where this fits in the GEO picture
A tool like this is where Google, Microsoft and Mozilla meet a coding agent on the way to the output. The three browser vendors are the people who decide what native means next year. Putting that decision into the agent before it writes code is faster than fixing stale code in review.
A team already using an agent to write front-end code can install Modern Web Guidance. Run the legacy-scan on one repository this week. If the scan returns findings the team has not seen in review, the pack is adding something review was missing. If it finds nothing, the team's review is already strong and the pack is a safety net. Either outcome tells the team what to change about how it buys agent tools next.
Source
- Modern Web Guidance 1.0, Chrome for Developers, Philip Walton
- GoogleChrome/modern-web-guidance, GitHub
- Google Releases Modern Web Guidance 1.0 For AI Coding Agents, Search Engine Journal, Matt G. Southern
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.