# Agent Browser

> Headless browser automation for AI agents

- Skill: `ericfeunekes/agent-browser` (Agent Skill)
- Install (CLI): `npx skillmds@latest add ericfeunekes/agent-browser`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ericfeunekes/agent-browser/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: ericfeunekes (https://skillmd.com/u/ericfeunekes)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/ericfeunekes/agent-browser

---


# Skill: `agent-browser` CLI

Headless browser automation for AI agents. Navigate, interact, screenshot, and extract content from web pages.

Run `agent-browser --help` for full command reference.

## Key patterns
- `agent-browser snapshot -i` — get interactive elements with refs (start here)
- `agent-browser click @e2` — interact by ref from snapshot
- `agent-browser get text @e1` — extract text content
- `agent-browser screenshot` — capture current state

## Ellis-specific flags (always include)
- `--json` — structured output where available

## Safety
- Do not submit forms, make purchases, or authenticate to services without user confirmation.
- Do not store or log credentials entered via `fill` or `set credentials`.
- Use `--session <name>` to isolate browser state between tasks.