agent-browser

Control browser sessions with isolated refs for deterministic web interactions.

40|9|Updated Jan 4, 2026
One-click install
npx skills add https://github.com/akillness/oh-my-skills --skill agent-browser-akillness
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-browser
Source: https://github.com/akillness/oh-my-skills/tree/main/.agent-skills/agent-browser
Command: npx skills add https://github.com/akillness/oh-my-skills --skill agent-browser-akillness

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Browser automation for AI agents can be brittle and hard to coordinate across tasks; this skill provides deterministic browser control to interact with web apps safely and reproducibly.

Core Features & Use Cases

  • Deterministic browser sessions with isolated refs for parallel tasks.
  • Open pages, interact with UI (click, fill, type, select, wait), and verify outcomes.
  • Capture evidence (screenshots, PDFs) and perform visual/content diffs for regression checks.
  • Use snapshot refs to keep actions deterministic across navigation and dynamic content.
  • Use cases include form automation, web testing, data extraction, and site verification.

Quick Start

Open a URL, wait for the page to be ready, and capture a deterministic snapshot to seed subsequent agent actions.

Frequently Asked Questions about agent-browser

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

FAQPage Schema
How do I automate browser interactions deterministically for AI agents?

Deterministic browser automation for AI agents is achieved by using isolated sessions and snapshot refs to interact with web pages. This approach ensures actions like clicking, typing, and waiting remain reproducible across dynamic content.

What is the best way to prevent cross-task leakage during parallel web automation?

To prevent cross-task leakage during parallel web automation, you need session isolation with isolated refs. This keeps multi-session orchestration separate, ensuring concurrent browser tasks do not interfere with each other.

How do I verify web app outcomes and capture evidence during headless browser testing?

Verifying web app outcomes in headless browser testing involves capturing screenshots and PDFs as evidence. You can also perform visual and content diffs for regression checks to confirm expected UI states.

Do I need a headless browser driver to run deterministic web automation?

Yes, a headless browser driver is required to run deterministic web automation. The skill relies on this driver to execute safe defaults, manage sessions, and control web interactions without a graphical interface.

Why does browser automation break with dynamic content and how do refs fix it?

Browser automation breaks with dynamic content because element locators shift during navigation. Using snapshot refs fixes this by seeding subsequent agent actions with deterministic anchors that remain stable across updates.

Can I use this browser automation skill for data extraction and form automation?

Yes, you can use this browser automation skill for data extraction and form automation. It enables agents to open pages, fill forms, select options, and verify outcomes deterministically across various websites and web apps.