UI Browser Test (localhost:8001)

SkillWeb & browsing

Runs quick UI smoke tests using Cursor's built-in browser MCP when the user asks to "启动内置浏览器测试UI/测试UI/用内置浏览器测一下". Navigate to http://localhost:8001 (port 8001) and validate basic UI functionality, console errors, and key network requests.

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the UI Browser Test (localhost:8001) skill

What this skill tells your AI

The instructions your AI receives, as published by hepai-lab/drsai in skills/skills_hepai/ui-browser-test-8001/SKILL.md and read by ahel’s review.

When to use

Use this skill when the user asks you to start Cursor's built-in browser to test UI, mentions "内置浏览器", "UI测试", "8001端口", or wants you to open and test the page on port 8001.

Default target

  • URL: http://localhost:8001

If the user specifies a different path (e.g. /chat) or query params, prefer the user-provided URL while keeping host/port on localhost:8001 unless explicitly changed.

Workflow

0) Preconditions

  • If the page cannot be reached (connection refused / blank), assume the dev server is not running and tell the user what you observed and what URL failed.
  • Do not guess alternative ports unless the user asks.

1) Open the built-in browser safely

Follow the built-in browser best practices:

  • Start by listing existing browser tabs. If a tab already exists, lock it before interacting.
  • If no suitable tab exists, navigate directly to http://localhost:8001, then lock the tab.
  • Between interactions that may change the DOM/URL, take a fresh snapshot before the next structural action.
  • Unlock only when completely finished with all browser operations.

2) Smoke-test checklist (fast, deterministic)

On the loaded page, perform a quick pass:

  • Visual verification: take one screenshot after load.
  • Console: check browser console messages; treat uncaught exceptions and React/Vue runtime errors as failures.
  • Network: inspect network requests; flag 4xx/5xx responses and failed fetch/XHR as failures.
  • Basic interaction (minimal):
    • If there is an obvious primary action (e.g. a main button, input, navigation tab), perform exactly one click or one short input action.
    • After the action, re-check console + network + take a second screenshot if the UI changed.

3) Report format

Return results in this compact structure:

  • URL tested: the final URL actually opened
  • Page state: loaded / partial / failed to load (+ error)
  • Console: OK / issues (paste the most important message lines)
  • Network: OK / issues (list failing request URLs + status)
  • UI checks:
    • What you clicked/typed (if any)
    • What changed (if anything)
  • Artifacts: screenshots taken (describe them; no embeds)

4) If failures happen

  • Prefer evidence: screenshot + console + failing network request.
  • Suggest the smallest next step: e.g. "start dev server", "fix 500 on /api/xxx", "missing env var", etc.

Signals

GitHub stars
24
Forks
5
Last commit
Sep 2026
Advanced
Item type
skill
Key
ui-browser-test-8001
Source
github.com/hepai-lab/drsai