agentleFS
bradbushSFAIbradbushSFAI/nolty/skills/chatgpt-image/SKILL.md

chatgpt-image

Skill1 starsChanged 6 months agoMITView on GitHub

This skill generates images using OpenAI's gpt-image-2 model. Use when users ask to generate photorealistic images, product photos, diagrams, mockups, brand-safe illustrations, or business graphics. Trigger phrases include "chatgpt image", "openai image", "photorealistic image", "product photo", "diagram", "mockup", "brand-safe image".

  • Reads credentials

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No trademark awareness - human review needed for logos

    line 30
  2. Very specific:

    When the user requests an image, run this Python script using uv run:

    line 34
  3. Distinctive:

    Check existing files to determine next number

    line 92
  4. Distinctive:

    Write the Python script to a temp file

    line 104

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

ChatGPT Image Generation Skill

Generate images using OpenAI's gpt-image-2 model (released April 2026) — optimized for photorealism, instruction-following, near-perfect text rendering, and business-ready assets. Uses O-series reasoning to research, plan, and self-check before rendering. Supports up to 2K natively and 4K with custom dimensions.

Setup

Requires OPENAI_API_KEY env var and uv (Python script runner). See docs/SETUP_OPENAI.md for the full walkthrough (API key creation, billing, env setup).

For Nolty to reach for this skill when the user asks "generate an image" or "draw X", make sure your TOOLS.md has the image-generation row pointing at this skill (the shipped TOOLS.template.md already includes it).

Model Strengths

Best for:

  • Photorealism - Skin, fabric, glass, metal render correctly. Lighting instructions stick.
  • Precise instruction following - Spatial layouts, cutaways, exploded views, ordered elements
  • Brand-safe illustration - Flat styles with palette discipline, cohesive icon sets
  • Text in images - Short headlines are legible, good for posters/covers/slides
  • Style transfer - Applies styles without breaking anatomy or subject identity
  • Business mockups - Stakeholder-ready visuals, faster than stock + Photoshop

Limitations:

  • Hands under extreme poses can fail
  • Tiny dense text on labels may drift
  • Long paragraphs of text will have errors
  • No trademark awareness - human review needed for logos

Usage

When the user requests an image, run this Python script using uv run:

# /// script
# requires-python = ">=3.11"
# dependencies = ["requests"]
# ///

import base64, os, requests, sys
from datetime import datetime

OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]

# Parse arguments
prompt = sys.argv[1] if len(sys.argv) > 1 else "A simple test image"
size = sys.argv[2] if len(sys.argv) > 2 else "1024x1024"
output_path = sys.argv[3] if len(sys.argv) > 3 else f"{datetime.now().strftime('%Y%m%d')}-openai-output.png"

print(f"Generating image with OpenAI gpt-image-2...")
print(f"Prompt: {prompt}")
print(f"Size: {size}")

resp = requests.post(
    "https://api.openai.com/v1/images/generations",
    headers={"Authorization": f"Bearer {OPENAI_API_KEY}"},
    json={
        "model": "gpt-image-2",
        "prompt": prompt,
        "size": size
    },
    timeout=120
)
resp.raise_for_status()
b64 = resp.json()["data"][0]["b64_json"]
with open(output_path, "wb") as f:
    f.write(base64.b64decode(b64))
print(f"Wrote {output_path}")

Size Options

Parse size from user request:

  • 1024x1024 - Square (default)
  • 1536x1024 - Landscape (wide)
  • 1024x1536 - Portrait (tall)
  • auto - Let model choose

User phrases to detect:

  • "landscape", "wide", "horizontal" → 1536x1024
  • "portrait", "tall", "vertical" → 1024x1536
  • "square" or no mention → 1024x1024

Output Naming

Marketing content mode (when working on a post file like 20251218 - Article Title.md):

  • Match the article naming: 20251218 - Article Title - 1.png
  • Increment number for additional images: - 2.png, - 3.png
  • Save in the same folder as the .md file
  • Check existing files to determine next number

General mode (standalone generation):

  • Format: YYYYMMDD-openai-[short-desc].png
  • Save in current working directory
  • Extract 2-4 word description from prompt
  • Example: 20251218-openai-raccoon-figurine.png

Workflow

  1. Parse the user's prompt and any size preferences
  2. Determine output path based on context (marketing article or general)
  3. Write the Python script to a temp file
  4. Run with uv run script.py "prompt" "size" "output_path"
  5. Display the generated image to the user

Example Prompts That Work Well

Product photography:

Studio product photo of a matte black ceramic raccoon figurine on a white-to-light-gray seamless background, softbox lighting, 85mm lens look

Diagrams:

Clean exploded view diagram of a smartphone showing internal components, technical illustration style, white background, labeled parts

Marketing graphics:

Flat illustration of a robot and human shaking hands, corporate blue color palette, minimal style, suitable for business presentation

Posters with text:

Modern conference poster with headline "AI Summit 2025" at top, abstract geometric shapes in blue and orange, professional corporate design

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.