browser

name	browser
description	Fetch and read web pages using Lightpanda headless browser, or automate full browser interactions using agent-browser. Use when you need to read articles, documentation, any web content, or interact with web pages.

Browser

Use this skill for fetching web content and automating browser interactions.

Shared guidance

• Always prefer Lightpanda for read-only fetching — it's faster and lighter.
• Use agent-browser when you need to interact with a page (click, fill forms, navigate, screenshot).
• Always set LIGHTPANDA_DISABLE_TELEMETRY=true when using Lightpanda.
• Always pass --obey_robots to Lightpanda (skip only for localhost).
• Fallback: if Lightpanda fails or crashes on a page, use agent-browser ($ab open '<url>' && $ab get text body) as a Chrome-based fallback.

Routing

Task	Reference to read
Read a web page, article, or docs	references/lightpanda.md
Interact with a page (click, fill, screenshot, test)	references/agent-browser.md
Authentication flows (login, OAuth, 2FA, state reuse)	{baseDir}/node_modules/agent-browser/skills/agent-browser/references/authentication.md
Multiple parallel sessions, state persistence	{baseDir}/node_modules/agent-browser/skills/agent-browser/references/session-management.md
Snapshot refs lifecycle, troubleshooting	{baseDir}/node_modules/agent-browser/skills/agent-browser/references/snapshot-refs.md
Video recording for debugging/docs	{baseDir}/node_modules/agent-browser/skills/agent-browser/references/video-recording.md
Performance profiling	{baseDir}/node_modules/agent-browser/skills/agent-browser/references/profiling.md
Proxy configuration	{baseDir}/node_modules/agent-browser/skills/agent-browser/references/proxy-support.md

Quick reference (read-only fetch)

For the most common case — just reading a page — use this directly:

LIGHTPANDA_DISABLE_TELEMETRY=true ~/.cache/lightpanda-node/lightpanda fetch \
  --dump markdown --strip_mode full --obey_robots '<url>'

For everything else, load the appropriate reference file.