analyzing-media
Analyzes PDFs, images, screenshots, diagrams, and documents using Gemini multimodal. Extracts text, tables, forms; interprets visuals, architecture diagrams, flowcharts, ERDs. Use when user mentions PDFs, images, screenshots, document extraction, OCR, visual analysis, diagram interpretation, or form processing. Do not use for web searching or shell commands.
When & Why to Use This Skill
This Claude skill leverages Gemini multimodal capabilities to analyze and interpret complex visual media, including PDFs, images, and technical diagrams. It excels at OCR, structured data extraction from forms and tables, and providing deep insights into visual architectures like flowcharts and ERDs, significantly streamlining document-heavy workflows.
Use Cases
- Automated Data Entry: Extracting structured information from scanned invoices, forms, and receipts to reduce manual input and errors.
- Technical Diagram Interpretation: Analyzing software architecture diagrams, flowcharts, or ERDs to generate documentation or explain system logic.
- Visual Content Search: Converting screenshots and images into searchable text and descriptive summaries for better knowledge management.
- Complex Document Analysis: Summarizing long PDF reports that contain a mix of text, charts, and tables for rapid information retrieval.
| name | analyzing-media |
|---|---|
| description | Analyzes PDFs, images, screenshots, diagrams, and documents using Gemini multimodal. Extracts text, tables, forms; interprets visuals, architecture diagrams, flowcharts, ERDs. Use when user mentions PDFs, images, screenshots, document extraction, OCR, visual analysis, diagram interpretation, or form processing. Do not use for web searching or shell commands. |
| allowed-tools | Task |
Gemini Media Analysis
Spawn the gemini-media-analyst agent for image and PDF analysis.
Foreground (blocking)
Task(subagent_type="gemini-media-analyst", prompt="Analyze @<file-path>: <what to look for>")
Background (for context efficiency)
Task(subagent_type="gemini-media-analyst", prompt="Analyze @<file-path>: <task>", run_in_background=true)
Use TaskOutput(task_id="<id>") to retrieve results.
Use for screenshots, architecture diagrams, PDFs, flowcharts, or any visual content analysis.