codex-vision

Route image review, generation, and editing tasks through OpenAI Codex.

Updated May 3, 2026
One-click install
npx skills add https://github.com/thenamangoyal/codex-vision --skill codex-vision
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: codex-vision
Source: https://github.com/thenamangoyal/codex-vision/tree/main
Command: npx skills add https://github.com/thenamangoyal/codex-vision --skill codex-vision

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

codex-vision removes the friction of routing image-related tasks through OpenAI Codex, so you can review an existing screenshot, generate a new visual, or edit an image using a single Claude Code shell skill.

Core Features & Use Cases

  • Codex-side screenshot review (review mode): Analyze an existing PNG/JPG/WebP/GIF and return Codex’s prioritized findings and fixes.
  • Image generation (generate mode): Produce a UI mock or visual artifact from a prompt via Codex image_gen, saving the result to an output path.
  • Image editing (edit mode): Modify an existing image by passing it to Codex with edit instructions, then save the edited output.
  • Session continuation controls: Optionally resume Codex sessions across follow-ups using cached UUIDs or explicit resume flags to keep context warm.

Quick Start

Use codex-vision to review a screenshot by running the review mode with an image path and your critique prompt.

Frequently Asked Questions about codex-vision

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

FAQPage Schema
How do I use Codex to review a UI mockup screenshot?

To review a UI mockup screenshot, route the image path and critique prompt through Codex using review mode. Codex analyzes the existing image and returns prioritized UX findings and suggested fixes.

Can I generate UI mockups with Codex from a text prompt?

Yes, you can generate UI mockups with Codex by using generate mode. This passes your prompt to Codex image_gen, which produces a visual artifact and saves the result to a specified output path.

How do I edit an existing image using Codex vision?

You can edit an existing image by invoking edit mode, which passes the image file and your edit instructions to Codex. Codex processes the modifications and saves the edited output image.

Does codex-vision support session continuation for follow-up image tasks?

Yes, codex-vision supports session continuation by resuming Codex sessions across follow-ups. It uses cached UUIDs or explicit resume flags to keep context warm for subsequent image generation or editing tasks.

What image formats can I send to Codex for screenshot critique?

You can send PNG, JPG, WebP, and GIF files to Codex for screenshot critique. The review mode accepts these formats to analyze existing images and return prioritized findings.

Do I need tmux to observe Codex image generation workflows?

Tmux is supported for observability during Codex image generation workflows but is not strictly required. It provides terminal session monitoring for review, generate, and edit modes when tracking Codex execution.