All skills
browserbase avatar

/ui-test

@b2ae728 official
by browserbasebrowserbase/skills3.7k stars
240

AI-powered adversarial UI testing via the browse CLI. Analyzes git diffs to test only what changed, or explores the full app to find bugs. Tests functional correctness, accessibility, responsive layout, and UX heuristics. Use when the user asks to test UI changes, QA a pull request, audit accessibility, or run exploratory testing. Supports local browser (localhost) and remote Browserbase (deployed sites).

Use this Skill: https://skilld.dev/gh/browserbase/skills/ui-test

This session only. Nothing lands on disk.

referencesexploratory-testing.md

≈740 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Exploratory Testing

Exploratory testing is agent-driven: you navigate the app freely using browse commands, making decisions about what to click, what to try, and what looks wrong — like a human QA tester.

How It Works

Unlike test suite execution (structured tests against specific routes), exploratory testing is open-ended. The agent uses browse snapshot to understand the page, makes a judgment call on what to do next, acts, and observes the result.

Workflow

1. browse open "TARGET_URL"
2. browse snapshot          → understand what's on the page
3. browse screenshot        → see what it looks like (Read the screenshot)
4. Decide: what looks wrong? What should I try?
5. browse click/fill/press  → interact
6. browse snapshot          → observe result
7. Repeat 4-6, navigating through the app freely

What to Look For

First Impressions (30 seconds)

  • Is the purpose of this page immediately clear?
  • Is there a clear visual hierarchy?
  • Does anything look broken, misaligned, or out of place?
  • Are there any console errors? (browse eval "JSON.stringify(window.__logs || [])")

Navigation Test

  • Can you reach every major section from the current page?
  • Does the back button work?
  • Are breadcrumbs accurate?
  • Are there any dead ends (pages with no way to navigate away)?
  • Does the logo link to the homepage?

Form Stress Test

  • Try submitting empty forms
  • Enter extremely long text (200+ characters)
  • Enter special characters: <script>alert('xss')</script>, "quotes", emoji 🎉
  • Double-click submit buttons rapidly
  • Tab through all fields — is the order logical?
  • What happens when validation fails? Is the error helpful?

State Persistence

  • Fill a form halfway, navigate away, come back — is data preserved?
  • Create an item, refresh the page — does it still exist?
  • Apply filters, refresh — are filters preserved?
  • Open a modal, press Escape — does it close cleanly?

Edge Cases

  • What does the page look like with no data (empty state)?
  • What happens when you navigate to a URL that doesn't exist (404)?
  • What happens with an expired session / no auth?
  • Resize the viewport to mobile — does it still work?

Performance Perception

  • Does the page feel fast or sluggish?
  • Are there loading indicators for slow operations?
  • Does content pop in or load smoothly?

Exploratory Testing Report Format

For each finding:

FINDING: [brief description]
SEVERITY: critical / high / medium / low
ROUTE: /path/where/found
EVIDENCE: [screenshot path or snapshot excerpt]
RECOMMENDATION: [specific fix suggestion]

Example:

FINDING: Settings page — "Regenerate API Key" button has no confirmation dialog
SEVERITY: high
ROUTE: /orgs/:slug/:projectId/settings/general
EVIDENCE: Clicked "Regenerate" and the key changed immediately with no warning
RECOMMENDATION: Add a confirmation dialog: "This will invalidate your existing key. Are you sure?"

Source: SKILL.md on GitHub

1 warning16d4 checks · Risk SAFE
  • Gen Agent Trust Hub16d

    The skill provides automated user interface testing scripts and recipes using the `browse` CLI tool. The dependencies and remote script sources specified are well-known and reputable technology entities, and the operations described strictly relate to its core purpose of visual and functional interface verification.

  • Socket16d

    No alerts

  • Snyk16d

    Risk: MEDIUM · 2 issues

  • ZeroLeaks5mo

    2 findings · Score: 80/100

Signed by skilld at b2ae728. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub last month.

Activeupdated 5 months ago
metadata
{
  "author": "browserbase",
  "version": "0.4.0"
}
All 1 allowed tools
Bash Read Glob Grep Agent
Other metadata
compatibility
Requires the browse CLI (`npm install -g browse`). For remote testing: BROWSERBASE_API_KEY and cookie-sync skill.

README badge

README badge for browserbase/skills/ui-test