Mini Browser (mb)

What it does

Mini Browser (mb) provides a unix-style CLI for browser automation that connects to Chrome over the DevTools protocol. It exposes small commands (go, text, shot, snap, fill, click, record, audit, js) which let an agent observe pages, interact with elements by coordinates or selectors, run JS, and produce screenshots or recordings.

When to use it

Use this skill when an agent needs to browse the web programmatically: automated testing, content scraping, visual audits, filling and submitting forms, capturing screenshots or screencasts, or running small JS snippets in page context. It's ideal for agents that prefer CLI-style primitives for navigation and interaction.

What's included

• Scripts: none shipped in the repo layer (has_scripts=false)
• References: none in the SKILL.md itself (has_references=false)
• Instructions: detailed command reference for navigation (mb go/url/back), observation (mb text/shot/snap), interaction (click/type/fill/key), recording, tab management, and an audit command (design/a11y/SEO). Setup and troubleshooting steps for Chrome/CDP are included.

Compatible agents

Best with code-capable agents that can call shell binaries and manage Chrome/CDP such as Copilot/Codex-style agents, Claude Code, or any agent with shell/CLI access and a headless/remote Chrome instance.