img-gen-avatar
Use when working on tools/img_gen avatar image generation, OpenAI-compatible image API config, human or yaoguai portrait prompts, qi-refining base generation, image-to-image realm edits, white-background postprocessing, manifests, or prompt rules that preserve pixel-art identity while changing cultivation realms.
DeepseekModel
官方收录技能
质量 优秀 · 90
v1.0.0
获取
https://deepseekmodel.com/api/download.php?id=4thfever-cultivation-world-simulator-cursor-skills-img-gen-avatar-skill-md&format=skill
下载 .skill
标准格式,含 system_prompt 与 model_config,导入任意 Agent 框架即可使用
.skill 文件中 system_prompt 字段的实际内容。
name img-gen-avatar description Use when working on tools/img_gen avatar image generation, OpenAI-compatible image API config, human or yaoguai portrait prompts, qi-refining base generation, image-to-image realm edits, white-background postprocessing, manifests, or prompt rules that preserve pixel-art identity while changing cultivation realms. Img Gen Avatar Core Files Read tools/img_gen/README.md for runnable commands. Read tools/img_gen/DESIGN.md before changing prompt structure, realm edit behavior, file naming, or postprocessing assumptions. Keep real API keys only in ignored tools/img_gen/image_api.env or environment variables. Never commit keys. Workflow Rules Use the two-stage avatar workflow: Generate only qi-refining base portraits with text-to-image. Generate foundation, golden core, and nascent soul portraits with image edits based directly on the qi-refining image. Do not use chained edits such as qi-refining -> foundation -> golden core -> nascent soul. Do not generate four-grid images. Default scripts must remain serial, skip existing files unless --overwrite is provided, show tqdm progress, and write manifests/failure JSON files. Prompt Rules Describe visible appearance only. Avoid invisible identity/background labels such as丹修、咒修、宗门、背景设定. For text-to-image base prompts, keep: pixel-art style low detail Q-style 2D anime portrait front-facing head-and-shoulders composition clean pure white background no texture, shadow, gradient, fog, glow, border, or scene background For image-to-image realm edits, preserve: input pixel-art style, low detail, and Q-style 2D anime look face shape, facial proportions, skin tone tendency, hairstyle, hair color eye color, pupil shape, expression baseline original facial marks, main accessories, subject size, head-shoulder ratio front-facing composition and pure white background For yaoguai edits, also preserve species traits and their positions/basic shapes, such as fox ears, wolf ears, ear feathers, vertical pupils, scales, turtle-shell patterns, and cheek markings. Realm edits may visibly strengthen gaze, pupil highlights, collar layering, accessory refinement, existing markings, and close-to-body glow. Keep the change obvious enough to show realm advancement, but never at the cost of pixel style or character identity. Explicitly forbid realistic style, high-definition illustration style, painterly rendering, over-detailed rendering, scene backgrounds, halos, frames, and large-area effects.
Agent 识别该技能的关键词,点击任意一个即可复制。
该技能未提供触发词。
下载的 .skill 包内含以下字段。
| 字段 | 说明 |
|---|---|
| format | 格式标识(skill/v1) |
| skill_id | 技能唯一 ID |
| name | 技能名称 |
| version | 版本号 |
| description | 技能描述 |
| category | 所属分类(数组) |
| trigger_words | 触发词列表 |
| tags | 标签列表 |
| source | 来源标识 |
| source_url | 来源链接(本页地址) |
| exported_at | 导出时间(每次下载生成) |
| system_prompt | 系统提示词正文 |
| model_config | 模型参数:provider / model / temperature / max_tokens / top_p |
| examples | 示例 |
| install_guide | 各平台导入说明(Coze / Dify / Claude / 自定义框架) |