10x Research-Skill by OpenAnalyst

Interactive content research, creation, web scraping, and AI image generation for Claude Code.

Features

• Web Research - Search and fetch content from the web
• Content Creation - Blogs, emails, social posts, web pages
• AI Image Generation - Gemini-powered image creation
• Web Scraping - Download all assets from websites
• Website Analysis - Analyze competitors and references
• Knowledge Base - Store and recall strategies, learnings, prompts
• Interactive Mode - Always asks before acting
• Smart Setup - Dependencies install only when needed

Quick Start

1. Setup

# Run setup (installs dependencies automatically)
python scripts/setup.py

Or use the command:

/setup

2. Configure API Key

Get your Gemini API key from: https://aistudio.google.com/apikey

Edit scripts/.env:

GEMINI_API_KEY=your-api-key-here

3. Use Commands

/help              # Show all commands
/research          # Research a topic
/create-content    # Create content
/generate-image    # Generate AI images
/scrape            # Download website assets
/analyze           # Analyze a URL
/plan              # Plan complex tasks
/explore           # Explore codebase/files
/execute           # Execute scripts/commands
/knowledge         # Manage knowledge base

Commands

Content & Research

Command	Description
/research [topic]	Research using WebSearch, WebFetch, download assets
/create-content [type]	Create blog, email, social, web page
/analyze [url]	Analyze website content and design

Media & Assets

Command	Description
/generate-image [prompt]	Generate AI images with Gemini
/scrape [url]	Download all assets from a website

Planning & Execution

Command	Description
/plan [task]	Plan and architect a task before execution
/explore [path/pattern]	Explore codebase, find files, search content
/execute [command]	Execute scripts, commands, and operations

Knowledge & Utilities

Command	Description
/knowledge	Manage knowledge base (strategies, learnings, prompts)
/setup	Initialize environment
/help	Show help

Natural Language

You can also use natural language:

• "Research AI trends" → Triggers /research
• "Create a blog post about..." → Triggers /create-content
• "Generate an image of..." → Triggers /generate-image
• "Scrape this website..." → Triggers /scrape
• "Analyze this competitor..." → Triggers /analyze
• "Plan this feature..." → Triggers /plan
• "Explore this codebase..." → Triggers /explore
• "Execute this script..." → Triggers /execute
• "Remember this..." / "Store this learning..." → Triggers /knowledge

How It Works

Interactive Mode

The AI always asks questions first before taking action:

User: Create a blog post

AI: I'll help create a blog post. Let me understand what you need:

1. What topic?
2. Should I research first? (WebSearch)
3. Target audience?
4. Tone/style?
5. Length?
6. Generate images?

Web Tools

Tool	Use Case
WebSearch	Find articles, trends, competitors
WebFetch	Read specific web pages
web_scraper.py	Download all assets locally
image_generator.py	Generate AI images

Workflow

CONTEXT (knowledge) → ASK → PLAN → SEARCH → DOWNLOAD → CONFIRM → CREATE → LEARN

Project Structure

Research-Skill/
├── .claude/
│   └── settings.local.json    # Claude settings, hooks
├── .claude-plugin/
│   └── plugin.json            # Plugin configuration
├── commands/                   # Slash commands
│   ├── research.md
│   ├── create-content.md
│   ├── generate-image.md
│   ├── scrape.md
│   ├── analyze.md
│   ├── plan.md
│   ├── explore.md
│   ├── execute.md
│   ├── setup.md
│   └── help.md
├── skills/                     # Skill definitions
│   ├── content-curator-workflow/
│   ├── image-generation/
│   └── web-scraper/
├── scripts/                    # Python scripts
│   ├── setup.py               # Environment setup
│   ├── image_generator.py     # AI image generation
│   ├── web_scraper.py         # Website scraping
│   ├── config.py              # Configuration
│   ├── organize_folders.py    # Folder management
│   ├── .env                   # API keys
│   └── utils/                 # Utility functions
├── knowledge/                  # Knowledge base (JSONL)
│   ├── strategies.jsonl       # Successful approaches
│   ├── plans.jsonl            # Workflow templates
│   ├── learnings.jsonl        # Key insights
│   ├── prompts.jsonl          # Effective prompts
│   ├── errors.jsonl           # Problems + solutions
│   └── templates.jsonl        # Content templates
├── data/                       # Generated data
│   ├── assets/                # Generated images
│   │   └── files/
│   └── scraped/               # Scraped websites
├── content-output/            # Created content
├── references/                # Reference files
│   ├── brand-voice.md
│   ├── audience-persona.md
│   ├── banner-guide.md
│   ├── blog-template.md
│   ├── email-template.md
│   └── scoring-template.md
├── CLAUDE.md                  # Claude instructions
├── DEMO.md                    # Demo prompts
├── README.md                  # This file
└── requirements.txt           # Python dependencies

Skills

1. Content Curator Workflow

Full content pipeline:

• Research with WebSearch
• Fetch content with WebFetch
• Score and rank sources
• Personalize with brand voice
• Create images
• Export in multiple formats

2. Image Generation

AI image generation with Gemini:

• Multiple styles (photorealistic, illustration, minimalist, etc.)
• Platform-optimized dimensions
• Automatic prompt enhancement

3. Web Scraper

Download website assets:

• HTML content
• Images (jpg, png, gif, webp, svg)
• CSS and JavaScript (optional)
• Videos
• Text extraction

Knowledge Base

JSONL-based storage for continuous learning:

Knowledge Types

Type	Purpose	File
Strategies	Successful approaches	strategies.jsonl
Plans	Workflow templates	plans.jsonl
Learnings	Key insights	learnings.jsonl
Prompts	Effective prompts	prompts.jsonl
Errors	Problems + solutions	errors.jsonl
Templates	Content templates	templates.jsonl

Usage

# Get context before a task
python scripts/knowledge_base.py context content

# Search knowledge
python scripts/knowledge_base.py search "image generation"

# Add new learning
python scripts/knowledge_base.py add learnings --title "..." --description "..."

# View statistics
python scripts/knowledge_base.py stats

Auto-Learning

After successful tasks, the system prompts to save:

• New strategies that worked well
• Effective prompts created
• Errors and their solutions
• Useful templates

Sub-Agents

Specialized agents for complex tasks:

Planning & Execution Agents

Agent	Type	Purpose
planner	Plan	Design implementation strategy
explorer	Explore	Fast file discovery and codebase exploration
executor	Bash	Command execution and terminal operations

File Operation Agents

Agent	Type	Purpose
reader	general	Read and analyze files
writer	general	Write and edit files

Content & Research Agents

Agent	Type	Purpose
researcher	general	Web research and content fetching
content-creator	general	Content writing and formatting
image-generator	general	AI image creation
scraper	general	Website asset downloading
analyzer	general	Website analysis

Hooks & Triggers

Natural Language Patterns

research|find|search|look up     → /research
create|write|make content        → /create-content
generate|create image|banner     → /generate-image
scrape|download|fetch site       → /scrape
analyze|check|review site        → /analyze
plan|design|architect            → /plan
explore|discover|browse          → /explore
execute|run|start                → /execute
knowledge|remember|learn|store   → /knowledge

On Init Hook

Automatically checks setup status on startup.

Smart Setup

The setup script only installs dependencies when needed:

python scripts/setup.py           # Install if needed
python scripts/setup.py --check   # Check status
python scripts/setup.py --force   # Force reinstall

Uses hash-based detection - only reinstalls when requirements.txt changes.

Configuration

Settings (.claude/settings.local.json)

• Permissions for web tools
• Hooks for automation
• Plugin configuration

Environment (scripts/.env)

GEMINI_API_KEY=your-api-key

Plugin (plugin.json)

• Command definitions
• Hook patterns
• Sub-agent configuration
• Reference file locations

Dependencies

google-genai>=1.0.0      # Gemini AI
pillow>=10.0.0           # Image processing
requests>=2.28.0         # HTTP requests
beautifulsoup4>=4.11.0   # HTML parsing
pyyaml>=6.0              # YAML support

Examples

Research a Topic

/research AI career trends

AI asks: How many sources? Download images? Timeframe?

Create Content

/create-content blog post about remote work

AI asks: Research first? Target audience? Tone and length?

Generate Image

/generate-image modern office workspace --style corporate --platform linkedin

AI asks: Confirm description? Any specific colors?

Scrape Website

/scrape https://example.com

AI asks: Download everything or images only? Limit?

Reference Files

File	Description
brand-voice.md	Brand tone guidelines, magic phrases
audience-persona.md	Target audience profile
banner-guide.md	Image design specifications
blog-template.md	Blog post structure
email-template.md	Email templates
scoring-template.md	Content scoring criteria

Scoring System (Content Curation)

Criteria	Points
Audience Relevance	30
Emotional Resonance	25
Actionable Value	20
Freshness	15
Authority	10

Golden Rules

1. Never assume - Always ask the user first
2. Get context first - Check knowledge base before tasks
3. Plan before execute - Use planning for complex tasks
4. Use web tools - Real research with WebSearch/WebFetch
5. Download assets - Save images and content locally
6. Learn and remember - Store successful strategies
7. Create what's requested - Match user's exact vision

Support

• See DEMO.md for demo prompts and examples
• Run /help for command reference
• Check CLAUDE.md for detailed AI instructions

---

10x Research-Skill v2.0 by OpenAnalyst - Interactive Content & Research for Claude Code