GPT Image 2 Search Skill by Image2Studio

Agent-ready prompt search for GPT Image 2, ChatGPT image generation, and Image2Studio visual references.

Agent skill for searching GPT Image 2 prompts, image references, and visual examples from Image2Studio.

This repository packages a single search skill for agents and Codex workflows that need reusable ChatGPT image generation prompts, example images, style references, composition patterns, or prompt inspiration. It is designed as a lightweight discovery layer before an agent writes, rewrites, compares, or adapts a GPT Image 2 prompt.

Included Skill

• gpt-image-2-search - search the public Image2Studio GPT Image 2 prompt library by concise visual keywords.

Use Cases

• Find prompt inspiration for a product shot, portrait, poster, scene, character, icon, UI image, or social media visual.
• Look up examples for a visual style, medium, mood, lighting setup, camera language, or composition.
• Collect reference images before writing or refining an image-generation prompt.
• Compare prompt structures and adapt a strong example to a new subject or brand direction.
• Recover from vague requests by searching for nearby visual language first.

Related Projects

• gpt-image-2-prompts - curated GPT Image 2 prompt gallery with 1,000 prompt-image pairs.
• Image2Studio prompt gallery - full searchable GPT Image 2 prompt library and web editor.

Install

Install only this skill:

npx skills add ryan-yang125/gpt-image-2-skills --skill gpt-image-2-search

Install the whole repository:

npx skills add ryan-yang125/gpt-image-2-skills

Quick Start

Run a local search:

node skills/gpt-image-2-search/scripts/search.mjs --q "cinematic portrait" --limit 3 --markdown

Search from inside the skill directory:

cd skills/gpt-image-2-search
node scripts/search.mjs --q "luxury perfume ad" --limit 5 --markdown

The command requires Node.js 18 or newer because it uses the built-in fetch API. No API key is required.

CLI Options

Option	Description
--q, -q	Search keywords. Required. Use short visual phrases.
--limit	Results per page. Defaults to 8, capped at 20.
--page	Page number. Defaults to 1.
--category	Filter by category.
--language	Filter by prompt language.
--featured	Filter featured prompts with true or false.
--api-base	Override the API base URL. Useful for testing.
--markdown	Print a compact Markdown summary instead of raw JSON.

Agent Workflow

1. Convert the user's request into 2-4 English visual keywords.
2. Search with node scripts/search.mjs --q "<keywords>" --limit 8 --markdown.
3. Check whether the results match the subject, format, style, lighting, composition, and intended use.
4. If the match is weak, retry once with broader or adjacent terms.
5. Present the strongest results with title, relevance, prompt excerpt, imageUrl, and studioUrl.

Good search phrases are concrete and visual:

cinematic portrait
minimal product poster
anime character sheet
luxury perfume ad
editorial fashion photo
surreal architecture

Output Fields

Search results may include:

Field	Purpose
title	Human-readable prompt title.
description	Short summary of what the prompt creates.
prompt	Reusable prompt text.
imageUrl	Reference image URL.
studioUrl	Web editor URL for opening the prompt.
categories	Prompt categories or tags.
stats	Weak popularity signal; do not rank only by this.

Project Structure

skills/
`-- gpt-image-2-search/
    |-- SKILL.md
    |-- references/
    |   `-- api.md
    `-- scripts/
        `-- search.mjs

Development

Run the smoke test:

node skills/gpt-image-2-search/scripts/search.mjs --q "cinematic portrait" --limit 3 --markdown

Validate a structured response:

node skills/gpt-image-2-search/scripts/search.mjs --q "cinematic portrait" --limit 2 \
  | node -e 'let s=""; process.stdin.on("data", d => s += d); process.stdin.on("end", () => { const j = JSON.parse(s); console.log({count: j.results.length, total: j.total, first: j.results[0]?.title}); });'