AgentSkillsCN

image-generation

利用 Gemini 进行 AI 驱动的图像生成与编辑。当 Claude 需要：(1) 根据文本描述生成图像；(2) 通过指令对现有图像进行编辑；(3) 制作信息图或图表；(4) 为报告或演示文稿生成视觉素材；(5) 处理 .png、.jpg、.jpeg、.webp 等格式的文件以进行编辑时，即可使用此功能。

SKILL.md
--- frontmatter
name: image-generation
description: "AI-powered image generation and editing using Gemini. Use when Claude needs to: (1) Generate images from text descriptions, (2) Edit existing images with instructions, (3) Create infographics or charts, (4) Generate visual assets for reports/presentations, (5) Work with .png, .jpg, .jpeg, .webp files for editing."

Image Generation & Editing

Overview

Generate and edit images using Gemini 2.5 Flash Image model via MCP tools.

Quick Start

Text-to-Image Generation

python
result = generate_image(
    prompt="A modern infographic showing renewable energy statistics",
    output_dir="/path/to/work_products/media"
)

Image Editing

python
result = generate_image(
    prompt="Change the background to a sunset over mountains",
    input_image_path="/path/to/original.png",
    output_dir="/path/to/work_products/media"
)

Preview Generated Image

python
# Auto-launch viewer after generation
result = generate_image(
    prompt="Blue square with rounded corners",
    preview=True  # Opens Gradio viewer
)

# Or preview any existing image
preview_image("/path/to/image.png")

Available Tools

generate_image

Primary tool for image generation and editing.

Parameters:

  • •prompt (required): Text description or edit instruction
  • •input_image_path (optional): Path to source image for editing
  • •output_dir (optional): Output directory (defaults to work_products/media/)
  • •output_filename (optional): Custom filename (auto-generated if not provided)
  • •preview (optional): Launch Gradio viewer after generation
  • •model_name (optional): Gemini model to use. Defaults to gemini-3-pro-image-preview.
    • •Options: gemini-3-pro-image-preview, gemini-2.5-flash-image.

Returns: JSON with output_path, description, size_bytes, and optionally viewer_url

describe_image

Get a short description of an image for generating meaningful filenames.

Parameters:

  • •image_path (required): Path to image file
  • •max_words (optional): Maximum words in description (default 10)

preview_image

Launch Gradio viewer to display an image.

Parameters:

  • •image_path (required): Path to image file
  • •port (optional): Port for Gradio server (default 7860)

Best Practices

Prompt Crafting

  • •Be specific: "modern, minimalist infographic with blue gradient"
  • •Include style: "photorealistic", "illustration", "line art", "cartoon"
  • •For charts/infographics: Include data points in the prompt
  • •For editing: Describe what to change AND what to preserve

Examples

Infographic:

python
generate_image(
    prompt="Infographic showing 3 key statistics: 45% growth, $2.3B market, 120 companies. Modern flat design with blue and green color scheme."
)

Chart:

python
generate_image(
    prompt="Clean bar chart comparing renewable energy adoption: Solar 35%, Wind 28%, Hydro 22%. Use professional colors, include labels."
)

Editing:

python
generate_image(
    prompt="Add snow to the mountain peaks, keep the sunset colors",
    input_image_path="work_products/media/sunset_mountains.png"
)

Integration with Other Skills

Reports (report-creation-expert)

Generate custom graphics for HTML reports:

python
# Generate infographic
img_result = generate_image(
    prompt="Infographic about AI market trends",
    output_dir="work_products/media"
)

# Embed in HTML report
html_content = f'''
<img src="file:///{img_result['output_path']}" alt="AI Trends" style="max-width: 100%;">
'''

PowerPoint (pptx skill)

Create custom slide backgrounds and visual elements:

  • •Generate images first, then reference paths in PptxGenJS
  • •Use addImage() to embed generated graphics
  • •Create infographics for data-heavy slides

Documents (docx skill)

Generate embedded figures and diagrams for Word documents.

Troubleshooting

Error: "GEMINI_API_KEY not set"

Ensure GEMINI_API_KEY is set in .env file.

Error: "No image data in response"

  • •Check your prompt is clear and specific
  • •Ensure Gemini 2.0 Flash model supports the request
  • •Try simplifying the prompt

Preview not working

  • •Gradio viewer script must exist at .claude/skills/image-generation/scripts/gradio_viewer.py
  • •Port 7860 must be available
  • •Run manually: python scripts/gradio_viewer.py /path/to/image.png

Output Conventions

  • •Directory: {workspace}/work_products/media/
  • •Filename pattern: {description}_{timestamp}.png
  • •Format: PNG (lossless)
  • •Description: Auto-generated from image analysis

Dependencies

  • •google-genai>=1.56.0 - Gemini API client
  • •gradio>=6.2.0 - Optional viewer UI
  • •pillow - Image processing