surf_skill

This skill lets you automate and inspect Chrome tasks from the CLI, enabling reliable browser testing, debugging, and AI-assisted workflows.

JavaScript

61

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 nicobailon/surf-cli --skill surf

SKILL.md16.7 KB

Overview

This skill exposes a battle-tested CLI to control Chrome for testing, automation, and debugging. It provides zero-config browser control for navigating pages, interacting with elements, taking screenshots, inspecting network/console, emulating devices, and issuing AI queries using logged-in browser sessions. The CLI is agent-agnostic and designed for fast, repeatable automation across development and QA workflows.

How this skill works

Commands talk to a running Chrome instance via CLI or Unix socket and use DevTools to drive navigation, input, screenshots, and network inspection. It reads page structure into element refs, supports semantic locating (role/label/text), and streams DevTools events for real-time debugging. AI queries run through the browser session so no API keys are required, provided you are logged into the target service in Chrome.

When to use it

Automating form fills, login flows, and repetitive UI tasks from scripts or CI
Capturing screenshots, full-page snapshots, or element-level images for visual tests
Inspecting network requests, console logs, and performance traces during debugging
Emulating devices, network/CPU conditions, geolocation, or touch input for QA
Running agent workflows that need controlled, isolated browser windows or tabs
Querying ChatGPT/Gemini/Perplexity/Grok via a logged-in Chrome session without API keys

Best practices

Call page.read first to generate stable element refs before interacting
Prefer semantic locate (role/label/text) over brittle selectors when possible
Use window isolation to create per-agent contexts and avoid cross-contamination
Validate AI UI bindings (e.g., surf grok --validate) when model selection fails
Record network or perf traces for intermittent issues and increase timeouts for slow models

Example use cases

Run a named workflow to log into an app, exercise features, and snapshot results for CI
Automate a multi-field form submission using form.fill or locate.label commands
Stream network events while reproducing a bug to capture failing requests and response bodies
Throttling network and CPU to test app behavior on slow mobile conditions
Ask Gemini or ChatGPT to summarize the current page context using --with-page for richer AI responses

FAQ

No. AI queries use the browser's logged-in session; you must be signed into the service in Chrome.

How do I handle UI changes that break commands?

Run validation (e.g., surf grok --validate) to inspect UI selectors, update local settings, or use semantic locate to target accessible roles and labels.

Can I run workflows in CI?

Yes. Workflows support dry-run validation, named JSON workflows, and can run headless or in isolated windows depending on environment setup.