Icon Designer Generator for Openclaw

1. Obtain a registered OneKey Gateway access key from the DeepNLP workspace.

export DEEPNLP_ONEKEY_ROUTER_ACCESS=your_access_key

2. Call the image generator through the OneKey Router. The minimum practical request includes a prompt, empty or uploaded images, a card count, the icon-designer template, and an asset size.

curl -X POST "https://agent.deepnlp.org/agent_router" \
  -H "Content-Type: application/json" \
  -H "X-OneKey: $DEEPNLP_ONEKEY_ROUTER_ACCESS" \
  -d '{
    "unique_id": "craftsman-agent/craftsman-agent",
    "api_id": "image_generator",
    "data": {
      "prompt": "Create a futuristic scientific icon about AI agents",
      "images": [],
      "card_count": 1,
      "template_id": "icon-designer",
      "asset_size": [{"width": 1080, "height": 1350, "ratio": "4:5"}],
      "mode": "basic",
      "styles": ["Futuristic scientific editorial style"]
    }
  }'

3. Alternatively, use the OneKey CLI after exporting the access key.

npx onekey agent craftsman-agent/craftsman-agent image_generator '{"prompt":"Create a futuristic scientific icon about AI agents","images":[],"card_count":1,"template_id":"icon-designer","api_id":"image_generator","asset_size":[{"width":1080,"height":1350,"ratio":"4:5"}],"mode":"basic","styles":["Futuristic scientific editorial style"]}'

No Node.js or Python package dependencies are listed in the skill manifest. Supported request fields include prompt, images, api_id, template_id, card_count, and asset_size; mode, styles, and design_config are optional.

Request fields

Design configuration taxonomy

• Profiles: channel_profile, audience_profile, age_band, profile_version, and resolved_profiles describe the intended audience and delivery channel.
• Generation controls: language, mode, text_model, and image_model control language and model selection; default values are supported.
• Creative direction: styles contains style blocks such as editorial, futuristic, or monochromatic treatments.
• Constraints: constraints captures limits such as maximum page count, density, or print requirements.
• Brief inputs: brief_inputs is an array of objects with id, type, and content; common types include prompt, style, and constraint.
• Text styles: text_styles can define brand, headline, content, and footer properties including color, font, size, weight, alignment, line height, and letter spacing.
• Page templates: page_templates defines reusable layers such as background, brand, image, text, and cta, each with position, size, and optional style overrides.

Response and generated files

Each page can include id, label, canvas_width, canvas_height, generation status, image prompts, content blocks, and editable layer values. Final assets can be exported from the workspace as PDF, PNG, or PPT.