Skillget

webmaton

agent
Security Audit
Warn
Health Pass
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 17 GitHub stars
Code Warn
  • network request — Outbound network request in skills/agentic-search/scripts/agentic_fetch.py
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Portable web-research and browser-automation SKILLs for Claude Code, Codex, and OpenCode — Playwright, Chrome DevTools, nodriver, and HTML-to-Markdown.

README.md

Webmaton

English | 中文

Acknowledgement

Special thanks to LINUX.DO — originally published on linux.do. Thank you to the community for the incredible support and feedback.

[!IMPORTANT]
These skills require uv and Python 3.13. Runnable skill entrypoints use PEP 723 inline metadata — dependencies resolve automatically via uv run. No requirements.txt needed.

Webmaton is a curated toolkit of portable, high-fidelity agent skills for web work — deep research, page capture, and browser automation. Each skill is self-contained, documented, and designed to drop into any agent runtime (OpenCode, Claude, Codex, and others) with minimal setup.

The name is a portmanteau of web and automaton — tools that let AI agents see, read, and interact with the web the way a human researcher would.

Skills

Skill What it does Best for
agentic-search Grok-primary deep research with grounded citations, Tavily/Firecrawl source discovery, verbatim extraction, and rerankable sessions. Research tasks that need sources, not summaries.
html-to-markdown Browser capture + deterministic HTML→Markdown conversion with metadata, link/image inventory, and quality signals. Converting JS-heavy pages or static articles into clean, structured Markdown.
nodriver-browser Persistent Chrome/Chromium automation via nodriver — clicks, typing, screenshots, DOM snapshots, and multi-step flows. Anything that requires interacting with a page like a human (logins, buttons, forms).
playwright-cli Playwright-backed browser sessions with snapshots, element refs, generated test code, storage, network, tracing, and video commands. Repeatable browser flows, Playwright test debugging, and test generation.
chrome-devtools-cli Chrome DevTools action CLI for snapshots, page interaction, console/network inspection, screenshots, Lighthouse, and performance traces. Frontend runtime debugging, layout inspection, and performance diagnostics.

Design principles

  1. Self-contained entrypoints — Every runnable skill entrypoint uses PEP 723 inline metadata, so dependencies resolve automatically via uv run. Private _*.py helper modules are imported by entrypoints and are not standalone commands.
  2. Composable sessions — agentic-search persists research sessions to disk, letting you search, extract quotes, rerank sources, and compose findings across multiple invocations.
  3. Browser-first fidelity — When a page needs JavaScript, login state, or DOM interaction, we reach for a real browser (Chrome → Chromium → Playwright fallback). For static content, we fetch directly. No overkill.
  4. Portable by default — Skills are symlink-friendly and runtime-agnostic. Drop them into ~/.codex/skills/, ~/.claude/skills/, or your agent workspace and they just work.

Quick start

Clone the repository and symlink the skills you need into your agent's skill directory:

git clone <repo-url> webmaton
cd webmaton

# Example: make agentic-search available to Claude
ln -s "$(pwd)/skills/agentic-search" ~/.claude/skills/agentic-search

Each skill's SKILL.md contains invocation examples, reference docs, and failure-mode guidance.

Requirements

  • Python 3.13
  • uv (for uv run script execution)
  • Node.js and npm for CLI-backed browser skills:
    • Node.js 18+ for playwright-cli
    • Node.js 20.19+ plus current Chrome stable for chrome-devtools-cli
  • API keys for providers you plan to use:
    • GROK_API_KEY / GROK_API_URL — for Grok-powered search and fetch
    • TAVILY_API_KEY — for Tavily search and site mapping
    • FIRECRAWL_API_KEY — for Firecrawl fallback fetching

Underlying tools

html-to-markdown uses markmaton for deterministic HTML-to-Markdown conversion, with nodriver handling browser-rendered capture when JavaScript is needed.

playwright-cli uses Microsoft's @playwright/cli. chrome-devtools-cli uses the chrome-devtools command from Google's chrome-devtools-mcp package.

License

MIT

Reviews (0)

No results found