safe-agentic-workflow: Command for Gemini CLI

description = "Analyze and describe images in a directory"

prompt = """

Analyze Images Command

Analyze images in the specified directory using Gemini's multimodal vision capabilities.

Target: {{args}}

Supported Formats

PNG, JPG, JPEG, GIF, WEBP, SVG, BMP

Workflow

1. Discover Images

Find all images in the target directory:

!{find {{args}} -type f \\( -iname "*.png" -o -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.gif" -o -iname "*.webp" -o -iname "*.svg" -o -iname "*.bmp" \\) 2>/dev/null | head -20}

2. Analyze Each Image

For each image found:

1. Read the image file using multimodal input
2. Describe the visual content
3. Identify key elements (objects, text, colors, composition)
4. Suggest appropriate filename based on content

3. Generate Report

Create a markdown report with:

## Image Analysis Report

### image1.png
- **Description**: [Visual description]
- **Key Elements**: [Objects, text, colors identified]
- **Suggested Name**: [content-based-name.png]
- **Dimensions**: [if available]

### image2.jpg
...

Output Options

Based on user request, can also:

• Rename files based on content
• Organize into category folders
• Extract text from images (OCR)
• Generate alt-text for accessibility
• Create thumbnails descriptions

Usage Examples

/media:analyze-images ./screenshots
/media:analyze-images ~/Downloads/images
/media:analyze-images ./assets/icons

Notes

• Gemini CLI multimodal: Images are base64-encoded for model consumption
• Large directories: Process in batches of 10-20 images
• Privacy: Image contents are sent to Gemini API for analysis """