mirror of https://github.com/NanmiCoder/cc-haha synced 2026-07-26 15:03:34 +08:00

feat(image-gen): add orchestration agent and prompt-craft skill (#64 )

The image-gen plugin previously shipped only the MCP server, leaving the
model with no behavioral guidance for when/how to call the tools. As a
result the agent often described what it would generate instead of
calling generate_image, and didn't know that returned image_url blocks
render inline automatically.

Add a plugin-scoped agent and skill following the spark2-gamedev
pattern (auto-discovered via agents/ and skills/ directories):

- agents/image-gen-agent.md: trigger rules, tool selection, prompt
  enhancement workflow, size inference, response patterns
- skills/prompt-craft/SKILL.md: 4 prompt templates (photorealistic,
  artistic, design, concept art), enhancement strategies per subject
  type (people/scenes/objects), zh→en translation guidance

The agent declares requiredMcpServers: [image-gen] so it only surfaces
when the MCP server is actually running.

Co-authored-by: 你的姓名 <you@example.com>

2026-06-17 00:24:27 +08:00

3.9 KiB

Raw Blame History

name	trigger	description
prompt-craft	When the user asks to generate, create, or design an image	Prompt engineering skill for image generation. Transforms short or vague user descriptions into rich, model-optimized prompts that produce high-quality results. Covers photorealistic, artistic, design, and concept art styles.

Prompt Craft Skill

将用户的简短描述转化为高质量的图像生成 prompt。

流程

分析用户意图（主题、风格、用途）
选择合适的 prompt 结构模板
扩展细节（光照、构图、氛围、材质）
添加质量修饰词
调用 generate_image 工具

Prompt 模板

写实摄影风格

[主体描述], [动作/姿态], [服装/外观细节],
[环境/背景], [时间/天气/氛围],
[灯光: 自然光/影棚光/黄金时刻/霓虹灯...],
[镜头: 85mm portrait lens / wide angle / macro...],
[摄影风格: editorial / street / fashion / documentary...],
highly detailed, professional photography, 8k resolution, sharp focus

示例转换：

用户："一个穿汉服的美女"
增强："A beautiful Chinese woman wearing traditional Hanfu dress, flowing silk in crimson and gold, elegant updo hairstyle with jade hairpin, serene expression, standing in a classical Chinese garden with blooming plum blossoms, soft golden hour sunlight filtering through bamboo, shallow depth of field, 85mm portrait lens, editorial fashion photography style, highly detailed, 8k, masterpiece"

艺术插画风格

[艺术类型: digital painting / watercolor / oil painting / anime / concept art...],
[主体描述], [动作/场景],
[色彩方案: warm palette / moody / vibrant / pastel...],
[艺术风格: Studio Ghibli style / art nouveau / cyberpunk aesthetic...],
[细节: intricate details / bold brushstrokes / cel shading...],
masterpiece, best quality, trending on artstation

设计/产品风格

[物体描述], [材质/纹理],
[背景: clean white / gradient / contextual environment],
[灯光: studio lighting / product photography lighting],
[视角: isometric / front view / 3/4 angle / flat lay],
professional product photography, clean composition, high resolution

概念艺术风格

[场景/世界观描述],
[建筑/环境元素],
[氛围: epic scale / mysterious / futuristic / ancient...],
[灯光/天气: volumetric lighting / god rays / storm clouds...],
[风格: concept art, matte painting, cinematic composition],
highly detailed environment, 4k, dramatic lighting

增强策略

人物类

用户输入	需要补充
只说了"人"	性别、年龄、表情、姿态、服装
只说了服装	人物特征、环境、光照
只说了风格	具体主题、构图、氛围

场景类

用户输入	需要补充
只说了地点	时间、天气、前景/中景/背景层次
只说了氛围	具体元素、色调、视角
只说了"风景"	具体地形、植被、光线、季节

物体类

用户输入	需要补充
只说了物品	材质、纹理、光照、背景、视角
只说了用途	具体设计风格、配色、上下文

多语言处理

用户中文输入 → 翻译为英文 prompt（大多数模型英文效果最佳）
保留文化特定术语的准确性（如"汉服"用"Hanfu"而非泛化为"Chinese dress"）
专有名词保持原样或使用公认英文对应词

尺寸推断

场景	推荐尺寸
头像、肖像、图标	1024x1024
全身人物、海报、手机壁纸	1024x1536
风景、场景、横幅	1536x1024
超宽电影感	1792x1024

注意事项

不要生成 NSFW 内容
不使用 negative prompt 语法（API 不支持）
prompt 控制在 50-200 词，过长反而降低质量
多次生成时适当变化构图和细节，而非重复同一 prompt

3.9 KiB Raw Blame History Unescape Escape

Prompt Craft Skill

流程

Prompt 模板

写实摄影风格

艺术插画风格

设计/产品风格

概念艺术风格

增强策略

人物类

场景类

物体类

多语言处理

尺寸推断

注意事项

3.9 KiB

Raw Blame History