Generate Image Skill
Generate images from natural language prompts using OpenAI's image API.
Script location
Installed by install.sh — path depends on your chosen install location.
Default: ~/.claude/skills/openai-image-skill/openai-image-skill.mjs
If the script is not found
If openai-image-skill.mjs is missing at the expected path, install it non-interactively:
git clone https://github.com/dkarski/openai-image-skill.git /tmp/openai-image-skill
bash /tmp/openai-image-skill/install.sh --default
rm -rf /tmp/openai-image-skill
Then retry the original image generation command.
When this skill applies
The user wants to generate, draw, or create an image — in any language.
How to invoke
Run the script with the Bash tool:
{{INVOKE_CMD}} "PROMPT" [--model=X] [--quality=low|standard|hd] [--size=X] [--count=N] [--output=X]
Extracting arguments from the user message
| User says | Map to |
|---|---|
| The description of what to draw | "PROMPT" (required, always quoted) |
| "using dall-e-2" / "model dall-e-3" | --model=dall-e-2 |
| "1792x1024" / "landscape" / "square" | --size=1792x1024 (use exact WxH) |
| "fast" / "cheap" / "draft" / "low quality" | --quality=low |
| "best quality" / "high quality" / "hd" / "detailed" | --quality=hd |
| "3 variants" / "give me options" / "multiple versions" | --count=3 |
| "save as cat.png" / "output: logo.png" | --output=cat.png |
If the user does not specify a flag, omit it — the script uses the saved default.
Size hints
- square →
1024x1024 - landscape →
1792x1024 - portrait →
1024x1792
Other commands
{{INVOKE_CMD}} list-models
{{INVOKE_CMD}} set-default dall-e-3
{{INVOKE_CMD}} update
What to report back
After the script exits successfully, tell the user:
- The saved file path (from
Saved: ...line in stdout) - The model used
- The size
If the script prints an error, show the error to the user and suggest the fix
(e.g. set OPENAI_API_KEY).
Example invocations
{{INVOKE_CMD}} "a futuristic city at night"
{{INVOKE_CMD}} "mountain landscape at sunset" --quality=hd
{{INVOKE_CMD}} "logo: minimalist fox" --model=dall-e-3 --count=3 --output=logo.png