Vision Multimodal

This adds Claude's vision capabilities to your code environment so you can process images, PDFs, and screenshots. You get base64 encoding patterns, multi-image comparison, and document extraction out of the box. The skill includes optimization helpers that resize images to around 1568px to cut token usage by 30-50%, which matters when you're processing multiple files. Works well for OCR-like text extraction, chart analysis, and structured data pulls from receipts or forms. One thing to note: it won't identify specific people and can struggle with handwriting, but for technical diagrams, UI screenshots, and printed documents it's solid.

skills.sh installs: 511
GitHub stars: 18
last updated: Jul 14, 2026

View on GitHub

Install

$ npx -y skills add lobbi-docs/claude --skill vision-multimodal --agent claude-code

Installs into .claude/skills of the current project.

Files

Vision & Multimodal Skill

Leverage Claude's vision capabilities for image analysis, document processing, and multimodal understanding.

When to Use This Skill

Image analysis and description
Document/PDF processing
Screenshot analysis
OCR-like text extraction
Visual comparison
Chart and diagram interpretation

Supported Formats

Format	Status	Best For
JPEG	✓	Photos, natural scenes
PNG	✓	Screenshots, UI, text
GIF	✓	Animated (first frame)
WebP	✓	Modern, compressed
PDF	✓	Documents (via Files API)

Image Size Guidelines

Minimum: 200 pixels (smaller = reduced accuracy)
Optimal: 1000x1000 pixels
Maximum: 8000x8000 pixels
Token cost: ~(width × height) / 1000
Tip: Resize to 1568px max dimension for 30-50% token savings

Core Patterns

Pattern 1: Single Image Analysis

import anthropic
import base64

client = anthropic.Anthropic()

# Load and encode image
with open("image.jpg", "rb") as f:
    image_data = base64.standard_b64encode(f.read()).decode("utf-8")

SKILL.md on GitHub