browser_skill

This skill automates browser tasks using agent-browser to open pages, snapshot accessibility refs, interact with elements, and verify UI across flows.

Python

458

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill gmickel/gmickel-claude-marketplace --skill browser

SKILL.md6.8 KB

Overview

This skill provides browser automation via the agent-browser CLI for navigating sites, interacting with UI, and capturing page state. It runs headless by default with an option for a headed window and uses accessibility snapshots to produce stable element refs for interactions. Use it to verify deployed UI, automate login flows, take screenshots, or scrape and inspect live pages.

How this skill works

The tool opens a URL, takes an accessibility snapshot to produce ref-based selectors (e.g. @e1), performs interactions using those refs or semantic finders, and re-snapshots after DOM changes to verify results. It supports waiting for network or JS conditions, parallel sessions, JSON output for machine consumption, and saving/restoring session state for auth reuse.

When to use it

Verify a deployed UI or capture baseline screenshots before a redesign
Automate end-to-end flows like login, form submission, and navigation
Scrape or extract structured page content for tooling or audits
Run quick smoke tests against a staging site or SPA after deploy
Inspect current page state, attributes, or accessibility tree

Best practices

Always run snapshot -i (interactive) to get stable ref tokens before interacting
Use wait --load networkidle or wait --fn for single-page apps to avoid race conditions
Re-snapshot after any DOM-changing action to refresh refs and verify results
Prefer semantic finders (role, text, label) when you know the element to reduce snapshot dependencies
Save session state after a successful login and reuse with state load for repeatable tests

Example use cases

Capture full-page and viewport screenshots of a landing page to compare design iterations
Fill and submit a signup form, wait for the dashboard URL, and snapshot to verify success
Run a smoke test suite: open site, verify key elements visible, check critical buttons enabled
Scrape product titles and prices using snapshot -i --json and parse get text results
Authenticate once, save auth.json, then run automated flows against authenticated endpoints

FAQ

Refs are stable tokens produced from the accessibility snapshot that map to DOM elements. Re-snapshot after any action that changes the DOM or navigation so refs stay accurate.

How do I test SPAs that load data asynchronously?

Use wait --load networkidle, wait --url patterns, or wait --fn with a JS condition to ensure the SPA has settled before taking snapshots or interacting.