install
source · Clone the upstream repo
git clone https://github.com/neojarvis0704/azure-image-gen
Claude Code · Install into ~/.claude/skills/
git clone --depth=1 https://github.com/neojarvis0704/azure-image-gen ~/.claude/skills/neojarvis0704-azure-image-gen-azure-image-gen-d3eea1
manifest:
SKILL.mdsource content
Image Generation & Editing Skill (Azure OpenAI)
This skill allows the agent to generate new images from text prompts or edit existing images using reference files via Azure OpenAI's GPT-Image-1.5 model.
Configuration
- Script:
gen.py - Environment:
(Required for API credentials)/home/thomas/.env.azure - Output: Images are saved to the
directory with a timestamped filename../out/
Usage
1. Generate New Image
Use this mode when the user wants to create an image from scratch. Syntax:
python3 gen.py "<detailed_prompt>"2. Edit Existing Image
Use this mode when a user provides one or more reference images and requests changes (e.g., changing colors, adding objects, or altering style). Syntax:
python3 gen.py "<edit_instructions>" --edit --ref <path_to_image>Parameters
| Parameter | Type | Description |
|---|---|---|
| String | (Positional) The visual description or edit instruction. |
| Flag | Required to trigger image-to-image/editing mode. |
| Path(s) | One or more paths to the images to be used as a reference. |
Examples
-
Create a new image:
python3 gen.py "A professional headshot of a solution consultant in a modern office, cinematic lighting" -
Modify an existing image:
python3 gen.py "Change the suit color to charcoal gray and add a laptop on the desk" --edit --ref ./out/20260208_120000.png
Notes for the Agent
- Fidelity: The skill is hardcoded to "high" fidelity to ensure identity and structural persistence during edits.
- Image Formats: Input images must be in PNG format.
- Output: The tool returns a JSON object containing the
to the generated image. Always display this path or the image to the user.path