Menu
Interactive web browsing and extraction with the Vercel Labs agent-browser CLI.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Update persistent swarm memory in ${SWARM_MEMORY_FILE} when the user explicitly asks to remember, update, or forget durable information.
Web search and content extraction via Brave Search API. Use for searching documentation, facts, or any web content. Lightweight, no browser required.
Author repo-root .forge project resources, including skills, specialists, reference docs, Project Agent definitions, Forge extensions, and Pi extensions/settings.
Use when creating, refining, or validating reusable global, project, or repository skills, including trigger wording, templates, helper scripts, and validation checks.
Test-only project skill fixture for validating repo-root .forge skill discovery. Use when the user asks to run the Forge resource smoke skill or validate project skill loading.
Create, update, list, and delete global slash commands.
| name | agent-browser |
| description | Interactive web browsing and extraction with the Vercel Labs agent-browser CLI. |
Use this skill for interactive browser tasks such as navigating pages, filling forms, extracting content, and taking screenshots.
Run the agent-browser CLI directly. This skill is documentation-only and does not provide wrapper scripts.
Global install (recommended):
npm install -g agent-browser
agent-browser install
Quick start without install:
npx agent-browser install
npx agent-browser open https://example.com
Project-local install:
npm install agent-browser
npx agent-browser install
agent-browser open https://example.com
agent-browser snapshot -i --json
agent-browser click @e2
agent-browser fill @e3 "hello"
agent-browser snapshot -i --json
agent-browser close
Use refs (@e1, @e2, ...) from snapshot output for deterministic interactions.
Navigation and session:
agent-browser open <url>
agent-browser back
agent-browser forward
agent-browser reload
agent-browser tab
agent-browser session
agent-browser session list
agent-browser close
Interaction:
agent-browser click <selector-or-ref>
agent-browser dblclick <selector-or-ref>
agent-browser focus <selector-or-ref>
agent-browser type <selector-or-ref> "<text>"
agent-browser fill <selector-or-ref> "<text>"
agent-browser press Enter
agent-browser hover <selector-or-ref>
agent-browser select <selector-or-ref> "<value>"
agent-browser check <selector-or-ref>
agent-browser uncheck <selector-or-ref>
agent-browser upload <selector-or-ref> "<path>"
agent-browser drag <source> <target>
agent-browser scroll down 500
agent-browser scrollintoview <selector-or-ref>
Semantic finding (no fragile CSS required):
agent-browser find role button click --name "Submit"
agent-browser find text "Sign in" click
agent-browser find label "Email" fill "test@example.com"
Waiting:
agent-browser wait 1000
agent-browser wait --load networkidle
agent-browser wait --text "Welcome"
agent-browser wait --url "**/dashboard"
agent-browser wait --fn "window.ready === true"
Extraction:
agent-browser get text <selector-or-ref>
agent-browser get html <selector-or-ref>
agent-browser get value <selector-or-ref>
agent-browser get attr <selector-or-ref> <attribute>
agent-browser get title
agent-browser get url
agent-browser get count <selector>
agent-browser snapshot -i --json
Capture and comparison:
agent-browser screenshot page.png
agent-browser screenshot --annotate
agent-browser screenshot --full fullpage.png
agent-browser pdf page.pdf
agent-browser diff snapshot
agent-browser diff screenshot --baseline before.png -o diff.png
agent-browser diff url https://v1.example https://v2.example
State and persistence:
agent-browser --session <name> open <url>
agent-browser --session-name <name> open <url>
agent-browser --profile <path> open <url>
agent-browser state save <path>
agent-browser state load <path>
agent-browser state list
Machine-readable output:
agent-browser snapshot --json
agent-browser get text @e1 --json
snapshot -i --json + refs (@eN) before interacting.snapshot after any action that may change the page.--session/--profile when isolation or persistence is needed.agent-browser close at the end of tasks to shut down the browser cleanly.