@ulpi/browse

browse skill

Foundational skill — teaches the agent the full browse CLI surface, the @ref workflow, and the target decision table.

Install

npx skills add https://github.com/ulpi-io/skills --skill browse

Trigger

Invoked automatically whenever the agent needs to browse a URL, test a web application, scrape content, or automate a native app. Also callable explicitly as /browse <request>.

What it does

The foundational skill for every other browse-based workflow. It teaches the agent:

  • The target decision table — when to use Browser vs Camoufox vs iOS Simulator vs Android Emulator vs macOS
  • The @ref workflow — every snapshot assigns refs (@e1, @e2…) that work as selectors in every command
  • All 128+ browse commands — navigation, content extraction, interaction, snapshots, visual, tabs, recording, performance audit, native app automation
  • Session patterns--session for parallel agents, --profile for persistent browser state
  • Safety and output flags--allowed-domains, --content-boundaries, --json, --max-output

Workflow pattern

Every browse task follows a predictable loop:

browse goto <url>            # Navigate
browse snapshot -i           # See interactive elements with @refs
browse click @e3             # Interact using refs
browse text                  # Read results

For native apps:

browse --platform ios --app com.apple.Preferences snapshot -i
browse --platform ios --app com.apple.Preferences tap @e1

When not to use

If a more specific skill applies (SEO audit → /browse-seo, stealth browsing → /browse-stealth, QA → /browse-qa), the agent should invoke that instead. The browse skill is the default/fallback.

On this page