nativewrightlisted

# NativeWright One principle: **route every browser automation action through the shared NativeWright daemon. Never use ad-hoc browser automation.** NativeWright wraps Patchright — a stealth-patched Playwright fork — in a long-lived HTTP daemon, so AI coding agents (Claude Code, Cursor, Codex, Gemini CLI) can drive a single real Chrome with a persistent profile across many one-shot shell calls. ## When to use this skill Use this skill when the task involves: - Opening any site that renders content via JavaScript (SPAs, React/Vue/Svelte apps). - Interacting with a real UI: clicking, typing, selecting, submitting forms, uploading files. - Visiting a site that requires login or session cookies. - Capturing screenshots, HTML, or DOM for evidence. - Stepping through a multi-step flow (checkout, wizard, generator). - Monitoring browser-side console or network traffic. ## When NOT to use Prefer lighter tools instead when: - The site returns the needed data in static HTML or a JSON API (use `WebFetch` or plain `curl`). - You only need a fact lookup (use `WebSearch`). - You are inspecting HTML files on disk (use `Read`). - You need to compare pure text content and the page has no JS (use `WebFetch`). Rule of thumb: launching a browser is expensive. Skip it when a plain HTTP request suffices. ## Locating the script The daemon self-registers its absolute path on every `start`. Read it from the platform-default install pointer: | OS | Install pointer | |---|---| | Windows | `%L