What is native-feel-skill?

native-feel-skill is a reference guide and "agent skill," a set of structured instructions you can install into an AI coding assistant, for building cross-platform desktop apps that look and behave like native apps rather than web apps wrapped in a browser. The problem it solves is the common tension between building once for multiple platforms (like macOS and Windows) and having the result feel polished and fast, not sluggish or "web-y." The skill was distilled from a technical deep-dive into how Raycast (a popular Mac productivity launcher) re-built its 2.0 version, plus reverse-engineering of the shipping Raycast Beta app. It captures eight architectural principles, a four-layer architecture (native shell, system WebView, Node backend, Rust core wired by a shared typed IPC contract), a survival guide for WebKit/WebView2 quirks (the embedded browser components used on each platform), and a 75-item checklist for auditing whether your app passes as native. You would use this when starting or refactoring a desktop app that must run on macOS and Windows from a shared codebase, supports a plugin ecosystem, needs to launch quickly and stay light on memory, and cannot sacrifice the feel of a real platform app. The skill is not intended for single-platform apps, games, or projects with very tight memory or startup budgets, the included decision-tree checklist explicitly rules itself out for those cases. Once installed, the skill activates automatically in conversation whenever cross-platform desktop architecture, WebView quirks, or native-feel questions come up. The full README is longer than what was provided.