Image Gen

Image Gen

Contents

• Overview
• When to Use This Skill
• Workflow
• Step 1: Analyze Article Content
• Step 2: Generate Cover Image
• Step 3: Generate In-Article Images
• Step 4a: Integrate Images
• Step 4b: Write ALT Text
• Step 5: Verify and Save
• Command Reference
• Prompt Engineering Best Practices
• Common Patterns
• Resources

Overview

Generate high-quality images for technical articles using the imagen CLI tool (powered by gemini-2.5-flash-image). This skill provides a systematic workflow for creating two types of images:

1. Cover Images: High-impact, conceptual images that grab attention on LinkedIn/Medium 2. In-Article Images: Technical diagrams and illustrations that explain specific concepts

The skill includes comprehensive prompt engineering guidance and best practices for integrating images with SEO-friendly ALT text.

When to Use This Skill

Invoke this skill when:

• Creating visual assets for a completed or near-complete article
• User requests "generate images for this article" or similar
• Article needs a cover image for publication
• Technical concepts would benefit from visual explanation
• Breaking up long text sections with relevant imagery
• Optimizing article for LinkedIn/Medium with visual appeal

Typical workflow position: After article content is finalized (Phase 8 in the article enhancement workflow), before SEO optimization.

Workflow

Step 1: Analyze Article Content

Before generating any images, analyze the article to identify:

1. Article topic and key themes: What is the main subject? 2. Key concepts needing visualization: Which technical ideas are complex or abstract? 3. Section breaks: Where would images improve readability? 4. Target audience: Developers? DevOps? Beginners or advanced?

Output: List of 2-4 images needed (1 cover + 1-3 in-article)

Example:

Article: "Mastering Alembic for Database Migrations"
Images needed:
- Cover: Conceptual image representing database evolution/migration
- Image 1: Diagram showing migration workflow
- Image 2: Visual comparing manual SQL vs Alembic automation

Step 2: Generate Cover Image

Cover images should be high-impact, conceptual, and visually interesting to grab attention. They are often metaphorical rather than literal.

Prompt Engineering for Covers:

• Style keywords: digital art, 3D render, conceptual art, abstract
• Atmosphere: cinematic lighting, dramatic, futuristic, professional
• Composition: Specify angle, focus, and elements
• Structure: [Style] cover image of [Metaphor/Concept] representing [Article Topic], [Atmosphere/Lighting], [Composition]

Command:

imagen generate "DETAILED PROMPT" --output work/images/[article_name]_cover.png

Example Prompts:

For a Vertex AI article:

imagen generate "A stunning digital art cover image showing three glowing pathways converging into a central AI cloud platform. Each pathway is a different color (green, blue, gold) representing different AI models flowing into Google Cloud infrastructure. The scene has dramatic cinematic lighting with a futuristic tech aesthetic. Style: conceptual, 3D render, high-detail, professional." --output work/images/vertex_ai_cover.png

For an Alembic article:

imagen generate "A dramatic 3D render showing a database schema transforming like evolving architecture, with migration files as glowing blueprints guiding the transformation. Dark background with cinematic lighting highlighting the evolution process. Style: conceptual art, digital art, high-detail, professional." --output work/images/alembic_cover.png

Tips:

• Be specific and detailed (100+ words is fine)
• Use metaphors that resonate with the topic
• Specify lighting/atmosphere for mood
• Include "professional" or "high-detail" for quality

Step 3: Generate In-Article Images (1-3)

In-article images should explain or reinforce specific concepts from the text. They are more direct and educational than cover images.

Prompt Engineering for In-Article Images:

• Style keywords: technical diagram, blueprint, minimalist illustration, clean
• Clarity: clear labels, connecting arrows, color-coded, organized
• Purpose: Focus on explaining one specific concept
• Structure: [Style] showing [Specific Concept], [Visual Elements], [Labels/Clarity]

Command:

imagen generate "CONCEPT-SPECIFIC PROMPT" --output work/images/[article_name]_[concept].png

Example Prompts:

For a technical diagram:

imagen generate "A clean technical diagram showing three distinct client libraries connecting to a central Vertex AI platform. Use color-coded arrows: green for native Google SDK, blue for partner Anthropic SDK, and gold for OpenAI-compatible SDK. Include clear labels for each connection path. Style: blueprint, technical diagram, clear labels, professional." --output work/images/vertex_ai_three_patterns.png

For a comparison illustration:

imagen generate "A minimalist side-by-side comparison showing traditional infrastructure (left) with tangled wires and manual configuration versus modern infrastructure-as-code (right) with clean, organized flow. Use contrasting colors and simple icons. Style: clean, minimalist illustration, symbolic." --output work/images/traditional_vs_iac.png

For a workflow diagram:

imagen generate "A flowchart-style diagram showing the Alembic migration workflow: 1) Model changes → 2) Auto-generate migration → 3) Review migration → 4) Apply to database. Use arrows connecting each step with clear labels. Style: blueprint, technical flowchart, organized, clean." --output work/images/alembic_workflow.png

Tips:

• Focus on one concept per image
• Use color-coding to distinguish elements
• Request "clear labels" or "connecting arrows" for clarity
• Simpler is often better for technical diagrams

Step 4a: Integrate Images

CRITICAL: Integrate generated images into the article at strategic locations.

Integration Checklist: 1. Create work/images/ directory if needed 2. Verify all images generated successfully (cover + in-article) 3. Add cover image immediately after H1 title 4. Add in-article images at strategic section breaks 5. Save updated article as new version (e.g., v8_with_images.md)

Markdown Integration Format:

# Article Title

![Cover image description here](images/article_cover.png)

## Introduction

Text introducing the problem...

![Diagram description here](images/article_diagram.png)

Step 4b: Write ALT Text

Write descriptive ALT text (50-125 characters) for each image to ensure accessibility and SEO.

Good ALT Text (descriptive, specific):

• "Architecture diagram showing Cloud Run connecting through VPC Connector to AlloyDB in private subnet"
• "Comparison showing manual infrastructure with tangled wires versus automated infrastructure-as-code"
• "Flowchart illustrating Alembic migration workflow from model changes to database update"

Bad ALT Text (vague, non-descriptive):

• "diagram" or "image"
• "architecture" (too vague)
• "image1.png" or "screenshot"

Step 5: Verify and Save

Final checks: 1. All images generated successfully (check file paths) 2. Images linked correctly in markdown 3. ALT text is descriptive and SEO-friendly (50-125 chars) 4. Cover image placed after H1 title 5. In-article images at logical section breaks 6. New version saved (e.g., work/draft/article_v8_with_images.md)

Typical outputs:

• 1 cover image (*_cover.png)
• 1-3 in-article images (*_concept.png, *_workflow.png, etc.)
• Updated article with integrated images and ALT text

Command Reference

Basic Usage:

imagen generate "PROMPT" --output FILENAME.png

Directory Setup:

mkdir -p work/images

Batch Generation Example:

# Cover
imagen generate "COVER PROMPT" --output work/images/article_cover.png

# In-article images
imagen generate "DIAGRAM PROMPT" --output work/images/article_diagram.png
imagen generate "COMPARISON PROMPT" --output work/images/article_comparison.png

Prompt Engineering Best Practices

For comprehensive prompt engineering guidance, refer to references/prompting_guide.md.

Core Principles: 1. Be Specific: Detailed descriptions produce better results 2. Define Style: Always specify artistic style (digital art, blueprint, etc.) 3. Set Atmosphere: Use lighting and mood keywords 4. Use Photography Terms: wide-angle, close-up, depth of field, etc.

Cover vs In-Article Summary:

Common Patterns

Pattern 1: Simple Article (1-2 images)

• Cover image only
• OR Cover + 1 key diagram

Pattern 2: Standard Article (2-3 images)

• Cover image (conceptual)
• 1-2 in-article images (technical diagrams)

Pattern 3: Complex Tutorial (3-4 images)

• Cover image (conceptual)
• Workflow/architecture diagram
• Comparison illustration
• Example/output visualization

Resources

references/prompting_guide.md

Complete guide to prompt engineering for the imagen CLI, including:

• Detailed prompt crafting principles
• Style and medium keywords
• Lighting and atmosphere techniques
• Photography/cinematography terms
• Extensive examples for different image types

Load this reference when crafting complex prompts or needing inspiration for visual concepts.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Repository Overview

This is a Claude Code skill for generating high-quality images for technical articles using the imagen CLI tool (powered by gemini-2.5-flash-image). The skill provides systematic workflows for creating both cover images and in-article illustrations with comprehensive prompt engineering guidance.

Core Commands

Image Generation

# Basic image generation
imagen generate "DETAILED PROMPT" --output FILENAME.png

# Example: Cover image
imagen generate "A stunning digital art cover image showing three glowing pathways converging into a central AI cloud platform. Dramatic cinematic lighting, futuristic tech aesthetic. Style: conceptual, 3D render, high-detail, professional." --output work/images/article_cover.png

# Example: Technical diagram
imagen generate "Technical diagram showing three color-coded client libraries connecting to central platform. Clear labels and arrows. Style: blueprint, technical diagram, clean." --output work/images/diagram.png

Directory Setup

# Create image output directory
mkdir -p work/images

Architecture and Workflow

Two-Phase Image Strategy

1. Cover Images: High-impact, conceptual images for grabbing attention on social platforms (LinkedIn/Medium)

• Style: digital art, 3D render, conceptual art, abstract
• Approach: Metaphorical and dramatic
• Keywords: cinematic lighting, futuristic, professional

2. In-Article Images: Technical diagrams and illustrations for explaining concepts

• Style: technical diagram, blueprint, minimalist illustration
• Approach: Direct and educational
• Keywords: clear labels, organized, clean

Standard Workflow Execution

1. Analyze Article: Identify key themes and concepts needing visualization 2. Generate Cover: Create one high-impact conceptual image 3. Generate In-Article Images: Create 1-3 technical diagrams/illustrations 4. Integrate with ALT Text: Add SEO-friendly descriptive ALT text (50-125 chars) 5. Save Updated Article: Version with integrated images

Prompt Engineering Principles

Core principles from references/prompting_guide.md:

1. Be Specific: Detailed descriptions (100+ words acceptable) 2. Define Style: Always specify artistic style explicitly 3. Set Atmosphere: Use lighting and mood keywords 4. Use Photography Terms: wide-angle, close-up, depth of field

Prompt Structure for Covers:

[Style] cover image of [Metaphor/Concept] representing [Article Topic], [Atmosphere/Lighting], [Composition]

Prompt Structure for Diagrams:

[Style] showing [Specific Concept], [Visual Elements], [Labels/Clarity]

SEO-Friendly ALT Text

Good Examples:

• "Architecture diagram showing Cloud Run connecting through VPC Connector to AlloyDB in private subnet"
• "Comparison showing manual infrastructure with tangled wires versus automated infrastructure-as-code"

Bad Examples (avoid):

• "diagram" or "image"
• "architecture" (too vague)
• "image1.png"

File Structure

image-gen/
├── SKILL.md                    # Skill definition and comprehensive documentation
├── CLAUDE.md                   # This file
├── README.md                   # User-facing documentation
├── references/
│   └── prompting_guide.md     # Detailed prompt engineering guide
└── scripts/                    # (Empty - placeholder for future automation)

Integration with Articles

Images should be integrated into markdown with this pattern:

# Article Title

![Conceptual cover image showing three glowing AI model pathways converging into Google Cloud platform](images/vertex_ai_cover.png)

## Introduction

Text introducing the problem...

![Technical diagram illustrating three-client pattern with color-coded SDK connections to Vertex AI](images/vertex_ai_three_patterns.png)

Common Patterns

Pattern 1: Simple Article

• Cover image only

Pattern 2: Standard Article

• Cover image (conceptual)
• 1-2 in-article images (technical diagrams)

Pattern 3: Complex Tutorial

• Cover image (conceptual)
• Workflow/architecture diagram
• Comparison illustration
• Example/output visualization

Important Notes

• The imagen CLI requires Python 3.12.9+ (managed via pyenv)
• All prompts should include explicit style keywords
• Cover images use metaphorical/conceptual approaches
• In-article images use direct/educational approaches
• Always provide descriptive ALT text for accessibility and SEO
• Save articles with version numbers when adding images (e.g., v8_with_images.md)

Reference Documentation

• SKILL.md: Complete skill documentation with workflow details
• references/prompting_guide.md: Comprehensive prompt engineering guide with examples and best practices

Image Gen - AI-Powered Article Image Generation

Generate high-quality cover images and technical diagrams for your articles using the imagen CLI tool powered by gemini-2.5-flash-image.

 

Overview

This Claude Code skill provides a systematic workflow for creating two types of images for technical articles:

1. Cover Images: High-impact, conceptual images that grab attention on LinkedIn/Medium 2. In-Article Images: Technical diagrams and illustrations that explain specific concepts

The skill includes comprehensive prompt engineering guidance and best practices for integrating images with SEO-friendly ALT text.

Installation

Universal Installer (Recommended)

The easiest way to install this skill is using the skilz universal installer:

skilz install SpillwaveSolutions_image_gen/image-gen

This will automatically download and configure the skill for use with Claude Code.

Manual Installation

Alternatively, you can manually clone the repository to your Claude Code skills directory:

git clone https://github.com/SpillwaveSolutions/image_gen.git ~/.claude/skills/image-gen

Prerequisites

• Python 3.12.9+ (managed via pyenv)
• imagen CLI tool installed
• Claude Code environment

Quick Start

# Generate a cover image
imagen generate "A stunning digital art cover image of [your concept]..." --output work/images/cover.png

# Generate a technical diagram
imagen generate "Technical diagram showing [your concept]..." --output work/images/diagram.png

Workflow

1. Analyze Your Article

Identify:

• Main topic and key themes
• Complex concepts needing visualization
• Optimal section breaks for images
• Target audience level

2. Generate Cover Image

Cover images should be conceptual and visually striking:

imagen generate "A stunning digital art cover image showing three glowing pathways (green, blue, gold) converging into a central AI cloud platform. Dramatic cinematic lighting, futuristic tech aesthetic. Style: conceptual, 3D render, high-detail, professional." --output work/images/article_cover.png

Style keywords: digital art, 3D render, conceptual art, cinematic lighting

3. Generate In-Article Images

In-article images should be clear and educational:

imagen generate "Technical diagram showing three color-coded client libraries connecting to central platform with clear labels and arrows. Style: blueprint, technical diagram, clean." --output work/images/diagram.png

Style keywords: technical diagram, blueprint, minimalist, clear labels

4. Integrate with ALT Text

Add images to your markdown with descriptive ALT text:

# Article Title

![Conceptual image showing three glowing pathways converging into cloud platform](images/cover.png)

## Section

Text content...

![Technical diagram illustrating three-client pattern with color-coded connections](images/diagram.png)

Good ALT Text (50-125 characters):

• "Architecture diagram showing Cloud Run connecting to AlloyDB via VPC Connector"
• "Comparison of manual infrastructure versus infrastructure-as-code automation"

Bad ALT Text:

• "diagram" or "image"
• "screenshot" or "image1.png"

Prompt Engineering Tips

For Cover Images

Goal: Grab attention with metaphorical, dramatic visuals

Structure:

[Style] cover image of [Metaphor/Concept] representing [Topic], [Atmosphere/Lighting], [Composition]

Example:

A dramatic 3D render showing a database schema transforming like evolving architecture, with migration files as glowing blueprints guiding the transformation. Dark background with cinematic lighting. Style: conceptual art, digital art, high-detail, professional.

For In-Article Images

Goal: Explain concepts clearly and directly

Structure:

[Style] showing [Specific Concept], [Visual Elements], [Labels/Clarity]

Example:

A clean technical diagram showing three distinct client libraries connecting to a central platform. Use color-coded arrows and clear labels for each connection path. Style: blueprint, technical diagram, professional.

Key Principles

1. Be Specific: Detailed descriptions produce better results (100+ words is fine) 2. Define Style: Always specify artistic style explicitly 3. Set Atmosphere: Use lighting and mood keywords 4. Use Visual Language: wide-angle, close-up, depth of field, connecting arrows

Common Image Patterns

Pattern 1: Simple Article (1-2 images)

• Cover image only
• OR Cover + 1 key diagram

Pattern 2: Standard Article (2-3 images)

• Cover image (conceptual)
• 1-2 in-article images (technical diagrams)

Pattern 3: Complex Tutorial (3-4 images)

• Cover image (conceptual)
• Workflow/architecture diagram
• Comparison illustration
• Example/output visualization

Complete Example

For an article on "Mastering Vertex AI":

# Step 1: Create output directory
mkdir -p work/images

# Step 2: Generate cover
imagen generate "A stunning digital art cover image showing three glowing pathways (green, blue, gold) converging into a central AI cloud platform representing Google Vertex AI. Dramatic cinematic lighting, futuristic tech aesthetic. Style: conceptual, 3D render, high-detail, professional." --output work/images/vertex_ai_cover.png

# Step 3: Generate technical diagram
imagen generate "Technical diagram showing three color-coded client libraries (green: Google SDK, blue: Anthropic SDK, gold: OpenAI SDK) connecting to central Vertex AI platform. Clear labels and arrows. Style: blueprint, technical diagram, clean." --output work/images/vertex_ai_diagram.png

# Step 4: Generate comparison
imagen generate "Side-by-side comparison: left shows traditional API integration with multiple adapters, right shows unified Vertex AI approach. Clean, minimalist. Style: minimalist illustration, professional." --output work/images/vertex_ai_comparison.png

Then integrate into markdown:

# Mastering Google Vertex AI: Three Client Patterns

![Conceptual image showing three glowing pathways converging into central AI cloud platform](images/vertex_ai_cover.png)

## Introduction

Problem statement...

![Technical diagram illustrating three client libraries with color-coded connections to Vertex AI](images/vertex_ai_diagram.png)

## Traditional vs Unified Approach

Comparison content...

![Comparison showing traditional multi-adapter API integration versus unified Vertex AI approach](images/vertex_ai_comparison.png)

Style Comparison

Additional Resources

• SKILL.md: Complete skill documentation with detailed workflow
• references/prompting_guide.md: Comprehensive prompt engineering guide with examples

Tips for Best Results

1. Be detailed in prompts - More description = better results 2. Specify style explicitly - Never assume a default style 3. Use color-coding - Helps distinguish elements in diagrams 4. Request clear labels - Essential for technical diagrams 5. Think metaphorically - For cover images, find visual metaphors 6. Keep it simple - One concept per image for in-article illustrations 7. Test and iterate - Generate multiple versions if needed

Links

• Skilz Marketplace: https://skillzwave.ai/skill/SpillwaveSolutions__image_gen__image-gen__SKILL/
• GitHub Repository: https://github.com/SpillwaveSolutions/image_gen

License

This skill is part of the Claude Code ecosystem.

Guide to Creating Compelling Article Images with Imagen

This guide provides instructions and best practices for using the imagen command-line tool to create high-quality images for technical articles, including cover images and in-article illustrations.

1. Basic Usage

The imagen tool is straightforward. The basic command structure is:

imagen generate "YOUR PROMPT HERE" --output FILENAME.png

• `imagen generate`: The command to start the image generation process.
• `"YOUR PROMPT HERE"`: A detailed description of the desired image. This is the most important part.
• `--output FILENAME.png`: The flag to specify the name of the output image file.

2. The Art of the Prompt: Core Principles

Image quality correlates directly with prompt quality. The model used (gemini-2.5-flash-image) is powerful, but requires clear instructions.

Principle 1: Be Specific and Detailed

Avoid vague terms. Instead of "a picture of a server," describe the server and its context.

• Simple: A server.
• Compelling: A single, sleek, modern server rack with glowing blue LED lights, standing in a clean, dark, and futuristic data center.

Principle 2: Define the Style and Medium

This is crucial for matching the tone of the article. Explicitly state the artistic style.

• Keywords: photorealistic, digital art, technical diagram, blueprint style, watercolor illustration, 3D render, minimalist line art, abstract, conceptual art.

• Example: A photorealistic image of... vs. A digital art illustration of...

Principle 3: Set the Scene with Lighting and Atmosphere

Describe the environment and the mood. Lighting is one of the most powerful tools for creating compelling images.

• Keywords: cinematic lighting, dramatic lighting, soft natural light, backlit, glowing, neon glow, dark and moody, bright and optimistic.

• Example: ...on a desk illuminated by the soft morning light from a window.

Principle 4: Use Photography and Cinematography Terms

Direct the "camera" to achieve the desired composition.

• Keywords: wide-angle shot, close-up macro shot, from a low angle, top-down view, portrait, landscape, depth of field, motion blur.

• Example: A close-up, macro shot of a single drop of water hitting a leaf.

3. Crafting Prompts for Article Images

For Cover Images

Cover images should be high-impact, conceptual, and visually interesting to grab the reader's attention. They are often more metaphorical than literal.

• Goal: Create a "wow" factor.
• Good Styles: digital art, 3D render, conceptual art, abstract.
• Prompt Structure: [Style] cover image of [Metaphor/Concept] representing [Article Topic], [Atmosphere/Lighting], [Composition].

Example Cover Prompt (for an article on AI ethics):

A stunning digital art cover image of a classic Greek statue holding a glowing, intricate neural network in its hands. The statue is cracked, showing circuits underneath. The lighting is dramatic, with a single spotlight on the statue in a dark room, representing the difficult choices of AI ethics. Style: conceptual, cinematic, high-detail.

For In-Article Images

These images should be more direct and serve to illustrate a specific point, break up text, and re-engage the reader.

• Goal: Explain or reinforce a concept from the text.
• Good Styles: technical diagram, blueprint, minimalist illustration, photorealistic (for concrete examples).

Example In-Article Diagram (for explaining a process):

This is the prompt we used successfully. It's a perfect example of a clear, technical request.

A detailed system architecture diagram for a URL shortening service similar to bit.ly. The diagram should show components like a load balancer, web servers, an application layer with a hash generation service, a distributed NoSQL database for storage, and a caching layer like Redis for fast lookups. Use clear labels and connecting arrows to show the flow of data. Style: clean blueprint, technical diagram.

Example In--Article Illustration (for an abstract point):

A minimalist illustration showing a small sapling growing out of a discarded computer keyboard, symbolizing the growth of new ideas from old technology. Style: clean, simple color palette, symbolic.

Combining these principles transforms simple pictures into compelling visual aids that elevate technical articles.