skills.rest
vZulin
vZulinCOMMUNITY
Explore all skills by this author →
GitHub

skyvern

Automates web tasks via LLM-driven visual UI navigation and interaction.

Updated Jun 3, 2026
Search Tags:#credentials#workflows#data-extraction#browser-automation#form-filling#vision-llm
One-click install
npx skills add https://github.com/vZulin/overlay-clock-timer --skill skyvern-vzulin
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: skyvern
Source: https://github.com/vZulin/overlay-clock-timer/tree/main/.codex/skills/macos-design/skyvern
Command: npx skills add https://github.com/vZulin/overlay-clock-timer --skill skyvern-vzulin

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Skyvern provides an AI-powered bridge to automate browser tasks using natural language, eliminating brittle selectors and manual clicks for repetitive web interactions.

Core Features & Use Cases

  • Natural language actions: click, type, scroll, hover, drag-and-drop, and form interactions through plain language.
  • Reusable workflows: build multi-step automation blocks and compose complex tasks with consistent results using vision-based reasoning.
  • Secure login: credential vault with 2FA/TOTP support to access protected sites without leaking credentials.
  • Broad tooling integration: compatible with Claude Desktop/Code, Cursor, Windsurf, and Codex to span different AI environments.
  • Robust against UI changes: uses Vision LLMs and computer vision to navigate dynamic interfaces.

Quick Start

Navigate to the target site, sign in with stored credentials, and run a multi-step workflow to collect data.

Frequently Asked Questions about skyvern

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate browser tasks using natural language instead of brittle selectors?▼

Yes, secure login for browser automation is supported through a credential vault with 2FA/TOTP support, allowing access to protected sites without leaking credentials. This ensures multi-step workflows operate safely across websites requiring authentication.

What is the best way to extract data from websites with dynamic user interfaces?▼

To run multi-step workflows, navigate to the target site, sign in using stored credentials, and execute the composed automation blocks. The system applies vision-based reasoning to complete complex tasks like data extraction and form filling with consistent results.

Does this browser automation approach work with Claude Desktop and Cursor?▼

Yes, this approach works with Claude Desktop, Claude Code, Cursor, Windsurf, and Codex. It requires Vision LLM capabilities and a browser engine alongside secure credential vaults with MCP tooling to operate safely across these AI environments.

Why does my web form filling automation break when the website UI changes?▼

Form filling automation breaks because traditional selectors fail when UI elements shift. Using Vision LLMs resolves this by interpreting the visual layout and navigating dynamic interfaces, maintaining robustness against UI changes without requiring script updates.

Related Skills

Explore More Software Engineering Skills→
lespacemanlespaceman
4

execute-task

Execute browser tasks from natural-language descriptions and capture evidence traces.

Community
Advanced
DNYoussefDNYoussef
4

stream-chain

Chain sequential agent outputs into multi-step workflow pipelines.

Community
Intermediate
AvaTar-ArTsAvaTar-ArTs

setup-tooluniverse

Guides users through ToolUniverse setup and MCP integration across multiple AI clients.

Community
Advanced
nbiishnbiish

skyvern

Automate browser workflows using vision-based interaction and Playwright.

Community
Advanced