3.4 KiB
| description | argument-hint | allowed-tools | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Generate a professional image using Gemini Flash via Nano Banana MCP from free-form description | <description> |
|
Generate a professional image based on the user description: "$ARGUMENTS"
Follow these steps:
-
Configuration Check: First verify that Gemini API is properly configured using
mcp__nano-banana-mcp__get_configuration_status -
Context Discovery: Automatically search for and gather relevant context from the project files:
- Search for any mentioned characters in
/heroes/directory - Look for locations in
/locations/or/cities/directories - Find world information in
/worlds/directory - Identify items in
/items/directory - Check for creatures in
/creatures/directory - Find applicable settings in
/settings/directory
- Search for any mentioned characters in
-
Entity Analysis: Parse the description to identify:
- Characters mentioned (explicit or implied)
- Locations referenced
- Items involved
- Creatures present
- Scene context and actions
- Appropriate style settings to apply
-
Prompt Generation: Create a professional prompt following the guidelines in
@docs/prompting-guide.md:- Use narrative descriptions instead of keyword lists
- Apply appropriate template from the prompting guide based on content type
- Include world-specific style elements
- Reference character unique identifying features
- Integrate linked items and relationships naturally
- Specify lighting and atmosphere
- Include technical quality descriptors
- Maintain visual continuity with established universe
-
Settings Integration: Automatically apply appropriate settings based on context:
- Character clothing based on status, world, and scenario
- Transportation methods suitable for world and character
- Environmental styles matching location and world
- Weapon styles appropriate for character and scenario
- Magical effects suitable for character abilities
-
Image Generation: Use
mcp__nano-banana-mcp__generate_imagewith the professionally crafted prompt -
File Management: Upon successful generation:
- Save image to appropriate directory based on content:
- Character portraits →
/heroes/{character-name}/photos/ - Location shots →
/locations/{location-name}/images/or/cities/{city-name}/images/ - Item visualizations →
/items/{item-name}/images/ - Creature images →
/creatures/{creature-name}/images/ - Scene compositions →
/scenes/{scene-name}/images/
- Character portraits →
- Archive the successful prompt in corresponding
/prompts/directory - Update cross-references if new relationships are established
- Save image to appropriate directory based on content:
-
Documentation: Create a markdown file in
/prepared_prompts/directory documenting:- The final prompt used
- Applied settings and context
- Generated image path
- Validation results and suggestions
- Entity relationships discovered
Important:
- The prompt must be in English regardless of input language
- Follow exact templates and best practices from
docs/prompting-guide.md - Respect the project framework defined in
CLAUDE.md - Use
mcp__nano-banana-mcp__generate_imagetool for actual image generation - Ensure proper file organization according to project structure