New Pack: kai-browser-skill v1.1.0 - Best of Both Worlds Browser Automation #309
Replies: 2 comments
-
|
MCP token overhead has gotten crazy. I love Claude Skills. Just started using this: https://github.com/lackeyjb/playwright-skill It uses /tmp dir pretty well to isolate the temp files. |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
Update: pai-browser-skill is now at v1.2.0 in PAI v2.1.0! 🎉 This version includes all the best-of-both-worlds improvements plus additional enhancements. Closing this announcement as superseded by the v2.1.0 release. See: https://github.com/danielmiessler/PAI/releases/tag/v2.1.0 |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
The Problem with Browser Automation Today
AI agents need browser automation for verification, screenshots, form filling, and testing. The standard approach is Playwright MCP:
{ "mcpServers": { "playwright": { "command": "npx", "args": ["@playwright/mcp@latest"] } } }The cost: ~13,700 tokens loaded at startup, whether you use it or not.
The Solution: File-Based MCP
Inspired by Anthropic's engineering blog post on code execution with MCP, the kai-browser-skill takes a different approach:
Code-first execution - Pre-written scripts that execute existing code, not generate new code.
Best of Both Worlds
Result: 99%+ token savings while retaining full Playwright functionality.
What's in v1.1.0
Get the Pack
kai-browser-skill.md
The pack includes:
Would love to hear if anyone else has explored file-based MCPs for browser automation. What's your experience with MCP token overhead?
Beta Was this translation helpful? Give feedback.
All reactions