junrei-browser-test

name	junrei-browser-test
description	Launch Junrei's isolated development environment and verify its web UI in a browser. Use when changing React UI, CSS/layout, navigation, interactive controls, or API-backed screen behavior in this repository, and whenever a dev server or port must be chosen for browser testing — with whatever browser tooling the current agent environment provides.

Junrei Browser Test

Start the environment

Run pnpm dev from the repository root. Do not select ports manually and do not use pnpm start for isolated agent testing.

Read the launcher output and use its Web URL. pnpm dev assigns free API and Web ports independently, starting from 7868 and 5874, and configures Vite's proxy to the assigned API port.

Use pnpm start only for the normal fixed-port mode: API 7867 and Web 5873 by default, overridable with JUNREI_PORT and JUNREI_WEB_PORT. Do not hardcode ports anywhere else (including .claude/launch.json, whose entries fit only this fixed-port mode) — pnpm dev allocates isolated ports dynamically.

When to test

Run browser verification when a change affects visible UI, routes, selection, layout, responsiveness, or client/server integration. Skip it for isolated non-UI logic when unit tests cover the behavior.

Run the test

The workflow is the same whatever agent harness is running. Never skip verification because a familiar tool surface is missing — use whatever browser control the environment provides:

• If a dedicated verifier agent is available (preview-verifier at .claude/agents/preview-verifier.md), delegate to it: it owns browser-tool usage, follows the port contract above, and returns a compact verdict. Pass it the Web URL or tab id when the environment is already running.
• Otherwise drive the browser tools directly: reuse a running Junrei tab or server whatever its port, else start one per the contract above, and navigate to the printed Web URL. In-app browser surfaces bootstrapped through a REPL (e.g. browser:control-in-app-browser via mcp__node_repl__js) document their own API — follow that skill rather than duplicating it here.

Inspect the DOM for content and interaction state; take a screenshot only when layout or visual hierarchy is under test.

Keep the development process running while testing and stop it once the task is finished unless the user asks to keep it available. Never stop servers you did not start.