Skill Detail
Run deterministic browser steps with settled screenshots and event logs for agents
Use Agent Browser Protocol when an agent needs browser actions to resolve into stable step results, complete with screenshots and surfaced events, instead of racing an always-live browser session.
Browser AutomationMulti-Framework
Browser Automation
Multi-Framework
Security Reviewed
Tool match: agent-browser-protocol
⭐ 436 GitHub stars
⬇ 1.7k/wk npm
BSD-3-Clause license
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill run-deterministic-browser-steps-with-settled-screenshots-and-event-logs-for-agents
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Tools required
Chromium-compatible local environment, Node.js/npx for MCP mode, MCP-compatible client or HTTP caller, vision-capable agent/client
Install & setup
Run `npx -y agent-browser-protocol` for local REST control or add `npx -y agent-browser-protocol –mcp` as an MCP server in a supported client, then drive browser actions as settled step calls.
Author
The Red Six
Publisher
Organization
Last updated
Apr 14, 2026
Quick brief
Use Agent Browser Protocol when the job is to turn flaky browser automation into a step-by-step agent workflow. An agent can navigate, click, type, and inspect pages through a local MCP or REST surface where each action waits for a settled browser state, returns a screenshot and event log, and freezes time again before the next decision. That boundary is tight enough to be skill-shaped: the skill is deterministic browser stepping for agents, not a generic browser framework card and not just another product listing for a Chromium fork.