Lightpanda

Use this skill instead of Chrome/Chromium when you need token-efficient content extraction and LP native semantic/interactive data.

Quick Start

lightpanda-start
lightpanda-fetch "https://example.com"
lightpanda-structured "https://example.com"
lightpanda-interactive "https://example.com"
lightpanda-semantic "https://example.com"
lightpanda-stop

Exported Commands

• lightpanda-start — start local Lightpanda CDP server
• lightpanda-stop — stop server started by this skill
• lightpanda-fetch — convert URL to markdown via LP.getMarkdown
• lightpanda-structured — extract LP.getStructuredData
• lightpanda-interactive — extract LP.getInteractiveElements
• lightpanda-semantic — extract LP.getSemanticTree

Runtime Model

• Browser binary is provisioned inside the installed skill directory at runtime/browser/lightpanda.
• Node dependency playwright-core is installed in-skill (node_modules/).
• PID/log state is owned under runtime/state/ for safe reset/clean operations.

Native LP Commands

• LP.getMarkdown
• LP.getSemanticTree
• LP.getInteractiveElements
• LP.getStructuredData

Notes

• DuckDuckGo is generally safer than Google for automated browsing with Lightpanda.
• Keep one CDP connection per process; use separate ports for concurrency.