sofunny-image

When to Use

• The user wants to generate images, but the current client's main model is not suitable for direct image output
• The user wants to upload one or more reference images and generate new images combined with text
• The user wants to generate images via New-API's native Gemini interface instead of directly connecting to Google's official API
• The user prefers to reuse:
  • Process environment variables

  • ~/.sofunny-image.env

Configuration Sources

1. ~/.sofunny-image.env

2. Current shell environment variables

--base-url

--api-key

--model

~/.sofunny-image.env

• SOFUNNY_BASE_URL

• SOFUNNY_API_KEY

• SOFUNNY_MODEL

• SOFUNNY_BASE_URL=http://127.0.0.1:3000

• SOFUNNY_MODEL=gemini-3.1-flash-image-preview

Installation and Execution Entry

• It is recommended to create a soft link of the repository directory to:

  • ${CLAUDE_PLUGIN_ROOT}/skills/sofunny-image

• Executable script entry:

  • ${CLAUDE_PLUGIN_ROOT}/skills/sofunny-image/scripts/sofunny-image.js

Quick Usage

node ${CLAUDE_PLUGIN_ROOT}/skills/sofunny-image/scripts/sofunny-image.js \
  --prompt "Generate a minimalist 16:9 banana poster without any text."

node ${CLAUDE_PLUGIN_ROOT}/skills/sofunny-image/scripts/sofunny-image.js \
  --prompt "Keep the subject unchanged and change the background to a valley in morning fog." \
  --input /absolute/path/to/ref-1.png \
  --input /absolute/path/to/ref-2.jpg

node ${CLAUDE_PLUGIN_ROOT}/skills/sofunny-image/scripts/sofunny-image.js \
  --prompt "Generate a cyberpunk-style banana icon." \
  --aspect-ratio 1:1 \
  --image-size 2K \
  --output /tmp/banana-icon.png

Parameter Description

• --prompt

  : Required, instruction for image generation or editing

• --input

  : Can be passed repeatedly, one or more reference images

• --output

  : Optional, output file path

• --aspect-ratio

  : Optional, default is

  16:9

• --image-size

  : Optional, default is

  1K

• --model

  : Optional, defaults to reading configuration, fallback to

  gemini-3.1-flash-image-preview

• --base-url

  : Optional, overrides the base URL in configuration

• --api-key

  : Optional, overrides the token in configuration

Workflow

1. Collect the user's prompt, reference images, output path, and aspect ratio requirements.
2. Run

   scripts/sofunny-image.js

   .
3. The script will directly call:

   • {BASE_URL}/v1beta/models/{MODEL}:generateContent

4. The script extracts the last

   inlineData.data

   from the response and saves it as the final image locally.
5. Return the save path to the user.

Notes

• This skill does not include

  tools

  and

  thinking

  by default to avoid failures of the image model in tool mode.

• BASE_URL

  should be the root address of the service, do not manually add

  /v1

  .
• If no output path is specified, the script will save the image by default to:
  • Current working directory
• Only use

  SOFUNNY_*

  variables in

  ~/.sofunny-image.env

  to avoid inconsistent behavior caused by mixing old configurations.
• If Gemini returns multiple images in the same response, the script will only save the last one by default, and previous images are considered intermediate products.