77 lines
1.6 KiB
Markdown
77 lines
1.6 KiB
Markdown
## TASK: Structured Prompt Generation (JSON Only)
|
|
|
|
You are an AI system that outputs a SINGLE JSON object.
|
|
Any output that is not valid JSON is INVALID.
|
|
|
|
You MUST comply with:
|
|
- README.md
|
|
- prompts/constraints.md
|
|
- prompts/world-set.md
|
|
- prompts/styles.md
|
|
- prompts/scenes.md
|
|
|
|
---
|
|
|
|
## 1. Active Configuration (Read-Only)
|
|
|
|
ACTIVE_WORLD_ID: AIART_CORE
|
|
ACTIVE_WORLD_VERSION: 1.0
|
|
|
|
ACTIVE_STYLE_ID: FLAT_BASE_01
|
|
ACTIVE_SCENE_ID: NONE_00
|
|
|
|
---
|
|
|
|
## 2. Task Objective
|
|
|
|
Generate a **character full-body prompt** for image generation.
|
|
|
|
Target character:
|
|
- Name: 白鴉（はくあ）
|
|
- Purpose: Character introduction reference
|
|
- Background: None
|
|
|
|
---
|
|
|
|
## 3. Output Requirements (Strict)
|
|
|
|
You MUST output a JSON object with EXACTLY the following structure:
|
|
|
|
```json
|
|
{
|
|
"meta": {
|
|
"world_id": "AIART_CORE",
|
|
"world_version": "WORLD_VERSION: 1.0",
|
|
"style_id": "FLAT_BASE_01",
|
|
"scene_id": "NONE_00",
|
|
"task_type": "character_full_body"
|
|
},
|
|
"prompt_text": "人物紹介用全身画像"
|
|
}
|
|
## Task Type Definitions (Binding)
|
|
|
|
The value of meta.task_type determines the scope of prompt_text.
|
|
|
|
- character_full_body
|
|
- Single character
|
|
- Full body visible
|
|
- No background narrative
|
|
|
|
- scene_composition
|
|
- Environment and spatial composition
|
|
- Characters may be absent or generic
|
|
- No named characters unless explicitly allowed
|
|
|
|
- item_design
|
|
- Standalone designed object
|
|
- No human presence
|
|
- Focus on form and material
|
|
|
|
- prop_design
|
|
- Small utilitarian object
|
|
- Human-scale
|
|
- No decorative symbolism
|
|
|
|
If prompt_text violates the declared task_type,
|
|
the output is INVALID.
|