What file types are supported by this skill?

It supports a wide range of formats including images (JPG, PNG, WebP), PDFs, videos (MP4, MOV), audio (MP3, WAV), and standard text-based documents.

How does Look At save context tokens?

Instead of loading the entire raw file into Claude's context, Look At processes the file externally via Gemini and returns only the specific information requested, minimizing token overhead.

When should I use Look At instead of the standard Read tool?

Use Look At for any file requiring visual interpretation or structured extraction, such as screenshots, diagrams, and PDFs with tables, charts, or complex formatting.

Do I need a specific API key to use this skill?

Yes, you must set the GOOGLE_API_KEY environment variable to allow the script to communicate with the Gemini API.

Look At - Multimodal File Analysis

Name: Look At - Multimodal File Analysis
Author: edwinhu

byedwinhu

数据科学与机器学习

Analyzes and extracts structured data from multimodal files like PDFs, images, and diagrams using Gemini models to minimize context token usage.

关于

Look At is a high-efficiency Claude Code skill designed to process and interpret visual or document-heavy files without overwhelming the conversation context. By delegating file analysis to Google's Gemini 2.5 Flash Lite model, it can extract specific information, describe UI screenshots, and parse complex tables from PDFs while saving between 80% and 95% of context tokens compared to standard file reading methods. This makes it an essential tool for developers needing fast, cost-effective insights from media files, charts, and architectural diagrams during the coding process.

主要功能

Provides visual descriptions for UI screenshots and architecture diagrams
0 GitHub stars
Reduces context token usage by up to 95% through targeted extraction
Supports multimodal analysis for images, PDFs, videos, and audio files
Leverages Gemini 2.5 Flash Lite for rapid and low-cost processing
Converts complex PDF tables and charts into structured data

使用场景

Summarizing technical diagrams and data flowcharts into plain text for Claude
Analyzing application screenshots to identify UI components and layout structures
Extracting specific data points or JSON structures from multi-page PDF documents

关于

主要功能

Provides visual descriptions for UI screenshots and architecture diagrams
0 GitHub stars
Reduces context token usage by up to 95% through targeted extraction
Supports multimodal analysis for images, PDFs, videos, and audio files
Leverages Gemini 2.5 Flash Lite for rapid and low-cost processing
Converts complex PDF tables and charts into structured data

使用场景

Summarizing technical diagrams and data flowcharts into plain text for Claude
Analyzing application screenshots to identify UI components and layout structures
Extracting specific data points or JSON structures from multi-page PDF documents