page-agent

Control live webpages via natural-language instructions using an in-page JavaScript agent.

Updated May 26, 2026
One-click install
npx skills add https://github.com/ruiyangruiyi/hermes-agent --skill page-agent-ruiyangruiyi
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: page-agent
Source: https://github.com/ruiyangruiyi/hermes-agent/tree/main/optional-skills/web-development/page-agent
Command: npx skills add https://github.com/ruiyangruiyi/hermes-agent --skill page-agent-ruiyangruiyi

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

page-agent eliminates the need for users to manually click through complex web interfaces by letting them control a live webpage using natural language instructions.

Core Features & Use Cases

  • In-page GUI agent: Embeds a pure-JavaScript/TypeScript agent inside your own web application and drives the current page via DOM text (no screenshots, no headless browser).
  • Natural-language DOM actions: Executes instructions like clicking UI elements and filling inputs based on what it can read from the page.
  • Client-side copilot embedding: Lets you ship an AI copilot to your SaaS, admin panel, or legacy web app while keeping integration lightweight (single script tag or npm package).
  • Evaluate against different LLMs: Test behavior using any OpenAI-compatible endpoint (cloud or local such as Ollama), including Qwen/OpenRouter/OpenAI-compatible services.

Quick Start

Embed page-agent in your page, configure an OpenAI-compatible LLM endpoint, and then tell it to "click login and fill the username as John" to drive the UI.

Frequently Asked Questions about page-agent

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

FAQPage Schema
How do I control a webpage through natural language without building a custom UI automation workflow?

You can embed a client-side JavaScript GUI agent in your web app to control live pages through natural language instructions. The agent reads DOM text and executes actions like clicking elements without needing screenshots or headless browsers.

Can I use a local LLM endpoint like Ollama to drive DOM interactions on an admin panel?

Yes, DOM interactions can be driven by any OpenAI-compatible LLM endpoint, including local models like Ollama. The in-page GUI agent uses the /v1/chat/completions interface to process instructions and execute actions on your admin panel.

What is the best way to add an AI copilot to a legacy web application without server-side browser automation?

The best way to add an AI copilot to a legacy web app without server-side browser automation is embedding a pure JavaScript/TypeScript agent. It reads the current DOM as text and executes UI actions directly on the client side.

How do I integrate natural language UI control into a SaaS application using a single script tag?

You can integrate natural language UI control into a SaaS app by adding a single script tag or npm package. This embeds an in-page agent that reads DOM text and executes actions like form filling via an OpenAI-compatible LLM endpoint.

Does the in-page GUI agent rely on screenshots or visual processing to execute web actions?

No, the in-page GUI agent does not rely on screenshots or visual processing. It reads the live webpage purely as DOM text, interprets natural language instructions via an LLM, and executes actions directly within the current page context.

What limitations should I expect when using a DOM text-based agent for web UI control?

Since the agent reads pages as DOM text and avoids server-side browser automation, it cannot process visual layouts via screenshots and requires a client-side OpenAI-compatible LLM endpoint to function properly within the live page context.