traktor

# Traktor - web asset extraction skill v2.0

Follow these steps exactly when extracting website assets.

## Trigger

`/traktor <url1> [url2] [url3]...`

---

## Step 1: Parse URLs from arguments

Extract all URLs from the user's command. Store them in a list.

```text
URLS = [all URLs provided after /traktor]
SITE_COUNT = number of URLs
PROJECT_DIR = current working directory
```

---

## Step 2: Create folder structure

### If SITE_COUNT == 1

**RUN THIS BASH COMMAND:**
```bash
mkdir -p assets/{logos,icons,images,videos,fonts,backgrounds} full-site/{pages,assets,styles}
```

### If SITE_COUNT > 1

**For each URL, extract site name (e.g., "stripe.com" -> "stripe-com") and RUN:**
```bash
# Replace {site-name} with actual site name for each URL
mkdir -p assets/{site-name}/{logos,icons,images,videos,fonts,backgrounds}
mkdir -p full-sites/{site-name}/{pages,assets,styles}
```

**Example for 3 sites:**
```bash
mkdir -p assets/{stripe-com,vercel-com,linear-app}/{logos,icons,images,videos,fonts,backgrounds}
mkdir -p full-sites/{stripe-com,vercel-com,linear-app}/{pages,assets,styles}
```

---

## Step 3: Spawn extraction agents

**For EACH URL, call the Task tool with these EXACT parameters:**

```text
Tool: Task
Parameters:
  - subagent_type: "general-purpose"
  - run_in_background: true
  - description: "Extract {site-name} assets"
  - prompt: [USE THE AGENT PROMPT BELOW - replace variables]
```

### Agent prompt (copy this exactly, replace {variables})

> **NOTE:** This agent prompt uses `mcp__claude-in-chrome__*` tools which require the
> claude-in-chrome browser extension MCP server. These tools will not be available
> in environments without the browser extension installed.

```text
Extract ALL assets from {URL} with paranoid-level thoroughness. Miss nothing.

OUTPUT DIRECTORIES
- Assets: {PROJECT_DIR}/assets/{site-name}/ (or {PROJECT_DIR}/assets/ if single site)
- Full site: {PROJECT_DIR}/full-sites/{site-name}/ (or {PROJECT_DIR}/full-site/ if single site)

PHASE 1:...