skills.rest
xintuchain
xintuchainCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

Vision Sandbox

Execute Python code in Gemini's sandbox to analyze images and extract spatial coordinates.

Updated Aug 27, 2026
Search Tags:#vision#sandbox#visual-analysis#code-execution#ui-auditing#spatial-grounding
One-click install
npx skills add https://github.com/xintuchain/tongtong --skill vision-sandbox-xintuchain
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: Vision Sandbox
Source: https://github.com/xintuchain/tongtong/tree/main/skills/vision-sandbox
Command: npx skills add https://github.com/xintuchain/tongtong --skill vision-sandbox-xintuchain

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires google-genai, and includes scripts (resource) components.

What problem does it solve?

Vision Sandbox enables precise visual analysis by allowing code execution inside Gemini's native sandbox to verify spatial data and visual reasoning results from images.

Core Features & Use Cases

  • Spatial grounding: extract coordinates and bounding boxes from UI images.
  • Visual reasoning with code: run Python code to process visual inputs and verify results.
  • UI auditing: check layout, overlaps, and accessibility programmatically.

Quick Start

Run a vision task on an image using the CLI to identify visual elements and coordinates.

Frequently Asked Questions about Vision Sandbox

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I extract bounding boxes and coordinates from UI images for spatial grounding?▼

Spatial grounding extracts bounding boxes and coordinates from UI images by executing Python code inside Gemini's native sandbox, enabling deterministic verification of visual elements and precise layout analysis.

Can I run Python code to audit UI layouts and check for accessibility issues programmatically?▼

Yes, UI auditing programmatically checks layout, overlaps, and accessibility by executing Python code in Gemini's sandbox to process visual inputs and verify visual reasoning results.

Do I need a GEMINI_API_KEY to perform visual analysis with code execution?▼

Yes, visual analysis with code execution requires a GEMINI_API_KEY in the environment, uv for dependency management, and Python 3.11 to run the vision_executor script.

What is the best way to verify visual reasoning results from images deterministically?▼

Visual reasoning with code provides deterministic verification by running Python code in Gemini's native sandbox to process visual inputs, validate spatial data, and confirm analysis results from images.

Why use code execution in a sandbox for visual analysis instead of standard model inference?▼

Code execution in a sandbox enables deterministic verification of spatial data and visual reasoning results, overcoming non-deterministic limitations of standard model inference for precise UI auditing tasks.

Related Skills

Explore More Software Engineering Skills→
spoonbobospoonbobo

nano-banana-pro

Generate and edit images with the Gemini 3 Pro Image model via Python.

Community
Intermediate
smykla-skalskismykla-skalski
3

ocr-finder

Detects and extracts text from image files with bounding boxes and click coordinates using EasyOCR.

Official
Intermediate
ovachieverovachiever
52

google-gemini-api

Integrate the Google Gemini API with the @google/genai SDK for generation, multimodal processing, and function calling.

Community
Advanced
johnlindquistjohnlindquist
24

gemini-image

Analyze images and extract text using Gemini Pro's vision API.

Community
Basic