feat(skills): add browser-use and crawl4ai

skills/software-development/browser-use/SKILL.md

@@ -0,0 +1,107 @@
+---
+name: browser-use
+description: "Use Browser Use to command an AI agent to navigate and use any website — fill forms, shop, make bookings, scrape data on demand. Source: https://github.com/browser-use/browser-use (105K GitHub stars)"
+version: 1.0.0
+author: Hermes Agent
+platforms: [linux, macos]
+tags: [browser, automation, web, scraping, ai-agent]
+source: https://github.com/browser-use/browser-use
+install: pip install browser-use
+---
+
+# Browser Use
+
+Browser Use lets an AI agent use a web browser the same way a human does — it opens pages, clicks buttons, types, and fills forms. You describe the task in natural language, and it completes it autonomously.
+
+**Key stats:** 9.5M downloads/week, 105K GitHub stars
+
+## Core Use Cases
+
+- **Form filling** — automate login, registration, data entry
+- **Web scraping** — extract structured data from any website on demand
+- **Booking** — flight, hotel, appointment booking automation
+- **Shopping** — price monitoring, cart automation
+- **Research** — navigate multiple pages, extract key info, compile report
+
+## Installation
+
+```bash
+pip install browser-use
+```
+
+Requires Python 3.10+ and Chrome/Chromium installed.
+
+## Basic Usage
+
+```python
+from browser_use import Agent
+from langchain_openai import ChatOpenAI
+
+agent = Agent(
+    task="Go to example.com, find the contact form, and fill it with name=test, email=test@example.com",
+    llm=ChatOpenAI(model="gpt-4o"),
+)
+agent.run()
+```
+
+## With Ollama (local models)
+
+```python
+from browser_use import Agent
+from langchain_ollama import ChatOllama
+
+agent = Agent(
+    task="Find the cheapest flight from Sydney to Melbourne next Friday",
+    llm=ChatOllama(model="llama3.3"),
+)
+agent.run()
+```
+
+## With Claude via Anthropic API
+
+```python
+from browser_use import Agent
+from langchain_anthropic import ChatAnthropic
+
+agent = Agent(
+    task="Log into stake.com.au and check my account balance",
+    llm=ChatAnthropic(model="claude-sonnet-4-20250514"),
+)
+agent.run()
+```
+
+## Key Parameters
+
+| Parameter | Description |
+|-----------|-------------|
+| `task` | Natural language description of what to do |
+| `llm` | LangChain LLM instance (OpenAI, Anthropic, Ollama, etc.) |
+| `browser` | Optional browser config (headless, viewport, etc.) |
+| `max_steps` | Max steps before stopping (default: auto) |
+
+## Verification
+
+```bash
+python3 -c "from browser_use import Agent; print('browser-use OK')"
+```
+
+## Integration with Hermes Ecosystem
+
+Browser Use complements the existing `computer-use` skill:
+
+- `computer-use` — drives the **local macOS desktop** in background (cua-driver)
+- `browser-use` — drives a **headless browser** on any website autonomously
+
+Browser Use can handle websites that require JS rendering, form submissions, and multi-step flows that static scraping cannot.
+
+## Pitfalls
+
+1. **Requires Chrome/Chromium** — install via `brew install chromium` or download from google.com/chrome
+2. **API costs** — if using OpenAI/Anthropic API, costs accrue per page/click. Use Ollama for free local inference
+3. **Anti-bot detection** — some sites (Cloudflare, CAPTCHAs) will block automated browsers
+4. **Headless by default** — set `headless=False` for debugging to watch the browser work
+
+## Related Skills
+
+- `computer-use` — local macOS desktop automation (cua-driver)
+- `crawl4ai` — LLM-optimized web crawler for bulk scraping

skills/software-development/crawl4ai/SKILL.md

@@ -0,0 +1,80 @@
+---
+name: crawl4ai
+description: "Crawl4AI: LLM-friendly web crawler and scraper. Convert any website into clean data ready for AI models. Source: https://github.com/unclecode/crawl4ai (67.8K GitHub stars, 2M downloads/month)"
+version: 1.0.0
+author: Hermes Agent
+platforms: [linux, macos]
+tags: [web, crawler, scraper, llm, ai, data-extraction]
+source: https://github.com/unclecode/crawl4ai
+install: pip install crawl4ai
+---
+
+# Crawl4AI
+
+Crawl4AI is an open-source LLM-friendly web crawler and scraper. It converts any website into clean, structured data ready for AI models — handles JavaScript-heavy pages that most scrapers struggle with.
+
+**Key stats:** 67.8K GitHub stars, 2M downloads/month
+
+## Core Use Cases
+
+- **LLM data preparation** — crawl and extract clean markdown/HTML for RAG pipelines
+- **Research automation** — pull content from multiple sites for synthesis
+- **Bulk scraping** — extract data from thousands of pages efficiently
+- **JavaScript-heavy sites** — handles SPAs, dynamic content, lazy-loaded pages
+
+## Installation
+
+```bash
+pip install crawl4ai
+```
+
+Or with Docker:
+```bash
+docker pull unclecode/crawl4ai
+```
+
+## Basic Usage
+
+```python
+import asyncio
+from crawl4ai import AsyncWebCrawler
+
+async def main():
+    async with AsyncWebCrawler(verbose=True) as crawler:
+        result = await crawler.araw_crawl("https://example.com")
+        print(result.markdown)  # Clean markdown output
+        print(result.html)      # Cleaned HTML
+
+asyncio.run(main())
+```
+
+## Key Features
+
+| Feature | Description |
+|---------|-------------|
+| `markdown` | Returns clean markdown (best for LLMs) |
+| `html` | Returns cleaned HTML |
+| `metadata` | Returns page metadata |
+| `js_content` | Execute JS before extracting (for SPAs) |
+
+## Verification
+
+```bash
+python3 -c "from crawl4ai import AsyncWebCrawler; print('crawl4ai OK')"
+```
+
+## Integration with Hermes Ecosystem
+
+Crawl4AI is a **free self-hosted alternative** to Firecrawl and Tavily. Use it for bulk research, RAG data preparation, and web extraction that doesn't need managed APIs.
+
+## Pitfalls
+
+1. **Infrastructure** — you manage your own crawler instances
+2. **JS pages** — may need `js_code` parameter to wait for hydration
+3. **Respect robots.txt** — check before bulk crawling
+4. **Cloud API** — Crawl4AI Cloud in closed beta if you want managed option
+
+## Related Skills
+
+- `browser-use` — autonomous AI browser agent for interactive web tasks
+- `firecrawl` or `web_search` — quick single-page lookups