sudoprivacy/sudowork/image-generation

Image Generation Skill

Generate new images or edit existing images via the images API.

When to Use

Trigger this skill when the user asks to:

• Generate / create / draw / design a new image ("generate an image of a cat", "draw me a sunset")
• Edit / modify / alter an existing image ("change the background to white", "add a hat to the person")
• Transform a referenced image ("make the above image black and white", "修改上面的图片")

Steps

Step 1: Determine operation type

• Generation: User describes a new image to create (no source image referenced)
• Edit: User references an existing image to modify

Step 2: For edits — identify the source image

• If the user provides a local file path, use it directly.
• If the user provides an image URL (http/https), download it first:

``bash curl -sL "<url>" -o /tmp/source_image.<ext> ``

• If the user references a previous image without a path (e.g. "the above image", "上面的图"):

1. List files in the workspace matching common image extensions (*.png, *.jpg, *.jpeg, *.gif, *.webp, *.bmp)
2. Sort by modification time (most recent first)
3. Use the most recent image. If there are multiple candidates and it's ambiguous, ask the user which image they mean.

Step 3: Run the image generation script

Use the cross-platform Python helper by default. Use the shell helper only when Python is unavailable and bash is available.

For generation (new image):

1. Determine output path stem: run pwd first to get the session workspace path, then construct an absolute path with a concise name (e.g. /session/workspace/white_cat). If the user specifies a location, use that absolute path instead.
2. Execute (use timeout: 180000):

``bash python skills/image-generation/scripts/generate_image.py gen "<prompt>" "<absolute_filepath_no_ext>" [size] ` If python is not available but python3 is, use python3. On Unix-like systems with bash, this fallback is also supported: `bash bash skills/image-generation/scripts/generate_image.sh gen "<prompt>" "<absolute_filepath_no_ext>" [size] ``

For editing (modify existing image):

1. Determine output path stem: use the session workspace path from pwd with the source filename and a timestamp (e.g. /session/workspace/abc_1712123456). If the user specifies a location, use that absolute path instead.
2. Execute (use timeout: 180000):

``bash python skills/image-generation/scripts/generate_image.py edit "<prompt>" "<image_path>" "<absolute_filepath_no_ext>" [size] ` If python is not available but python3 is, use python3. On Unix-like systems with bash, this fallback is also supported: `bash bash skills/image-generation/scripts/generate_image.sh edit "<prompt>" "<image_path>" "<absolute_filepath_no_ext>" [size] ``

The script appends the correct extension (.png, .jpg, .webp) based on the output image data.

• size is optional, defaults to 1024x1024. Common sizes: 1024x1024, 1024x1536, 1536x1024.
• The script reads API credentials and image model from ~/.nexus/sudocode/sudocode.json automatically. No manual configuration needed.
• The image model is read from sudocode.json (set via Tools settings). Always run the script directly — it will report any configuration errors itself.
• Do not manually set, override, print, echo, or probe PROVIDER_BASE_URL, PROVIDER_API_KEY, IMAGE_MODEL, or SUDOCODE_CONFIG_PATH.
• Do not curl or test alternate image providers/endpoints. If the helper fails, report the helper's error message instead of switching providers.
• Do not put API keys, bearer tokens, or provider credentials in shell commands, inline scripts, logs, or chat output.

The script prints the saved image file path on success, or an error message on failure.

Step 4: Show the result

Display the image inline using markdown image syntax so it renders in chat:

![<concise description>](<absolute_image_path>)

The script prints the saved file path — use that as <absolute_image_path>.