审美助手
aesthetic-copilot
by chenchen913
Use when the user wants to generate high-fidelity PROMPTS for Text-to-Image models (Flux, Ideogram, Midjourney) based on vague layout/content descriptions.
安装
claude skill add --url github.com/openclaw/skills/tree/main/skills/chenchen913/aesthetic-copilot文档
Aesthetic Copilot (v3.4 - Integrated)
Overview
This skill acts as a Meta-Prompt Generator. It translates a user's vague idea into a professional, commercial-grade image generation prompt optimized for models like Flux, Ideogram, and Midjourney v6.
File Map (MANDATORY READ ORDER)
| File | Role | When to Read |
|---|---|---|
SKILL.md | This file — entry point | Always |
engine/README.md | Content Classification Logic | Step 1 (Analyze) |
styles/premium/master-collection.md | The Vault: Premium Styles | Step 2 (Style Selection) |
styles/apple-minimal.md | Apple Minimal detail spec | When apple-minimal or apple-pro is selected |
styles/neo-brutalism.md | Neo-Brutalism detail spec | When neo-brutalism is selected |
styles/warm-academia.md | Warm Academia detail spec | When warm-academia is selected |
styles/cyber-glass.md | Cyber Glassmorphism detail spec | When cyber-glass is selected |
styles/nature-organic.md | Organic Nature detail spec | When nature-organic is selected |
engine/style-mixer.md | Randomization & Conflict Logic | Step 3 (Dice Roll) |
engine/micro-innovation.md | Artistic Twists | Step 4 (Innovation) |
layouts/README.md | Layout Templates | Step 5 (Layout Selection) |
prompt-templates/*.md | Output Skeletons | Step 6 (Generate) |
The Workflow (Strict Execution Path)
Step 1 — Analyze & Classify Intent
Action: Read engine/README.md.
Apply its classification logic to extract:
- Keywords from the user's input
- Sentiment / Mood (Warm, Playful, Serious, Futuristic…)
- Industry (Tech, Fashion, Education, Food…)
- Attribute scores:
formal_level,color_temp,contrast,complexity
Use these attributes to drive all downstream decisions.
Step 2 — Select Base Style
Action: Read styles/premium/master-collection.md.
Match the attributes from Step 1 to the closest style ID.
- If the matched style is
apple-minimalorapple-pro→ also readstyles/apple-minimal.mdfor detail tokens. - If the matched style is
neo-brutalism→ also readstyles/neo-brutalism.mdfor detail tokens. - If the matched style is
warm-academia→ also readstyles/warm-academia.mdfor detail tokens. - If the matched style is
cyber-glass→ also readstyles/cyber-glass.mdfor detail tokens. - If the matched style is
nature-organic→ also readstyles/nature-organic.mdfor detail tokens. - Fallback: If no confident match, default to
apple-pro.
Step 3 — Roll the Dice (Mixer)
Action: Read engine/style-mixer.md.
Randomly select (do NOT default to first item):
- ONE Material Twist from Pool A
- ONE Lighting Modifier from Pool B
- ONE Composition Rule from Pool C
Apply the Harmony & Conflict Resolution rules before proceeding.
Step 4 — Inject Micro-Innovation
Action: Read engine/micro-innovation.md.
- Find the Input Category that matches the user's subject.
- Apply the corresponding Twist (not the standard depiction).
- Determine the Text Integration method (Embossed / Neon / Integrated / Masked).
Step 5 — Select Layout
Action: Read layouts/README.md.
Match the user's description to a Layout ID:
| User Intent | Layout ID |
|---|---|
| "Top banner + columns", info layout | hero-split |
| "Left menu / sidebar" | sidebar-fixed |
| "Pinterest style", photo wall | masonry-grid |
| "Magazine cover", event poster | poster-zine |
| "Phone app", Instagram feed | mobile-feed |
Pass the selected Layout ID into the template as [Layout].
Step 6 — Route to Template
Select the correct template based on the classified intent from Step 1:
| Intent | Template |
|---|---|
| Magazine / Fashion / Editorial | prompt-templates/editorial-spread.md |
| Product / Object / Commercial | prompt-templates/product-showcase.md |
| Dream / Abstract / Surreal | prompt-templates/surreal-concept.md |
| Informational / Default | prompt-templates/structural-poster.md |
Action: Read the selected template file, then fill it with all values accumulated in Steps 1–5.
Output Format
Return the final result wrapped in a single plaintext code block:
**🎨 Aesthetic Copilot: Generated Prompt**
> **Template**: [Selected Template Name]
> **Style DNA**: [Base Style] + [Material Twist] + [Lighting Modifier]
> **Layout**: [Layout ID] — [Layout Name]
> **Concept**: [One sentence explaining the Micro-Innovation twist]
```
[Filled prompt content from the selected template]
```
Anti-Patterns
- Do not use
structural-posterfor everything — route correctly in Step 6. - Do not pick the first item in the Mixer pools — randomize.
- Do not skip Step 1 — content classification drives all downstream choices.
- Do not skip Step 5 — layout selection must be passed into the template.
- Do not use vague style guesses — always read
master-collection.mdfirst.
相关 Skills
Claude接口
by anthropics
面向接入 Claude API、Anthropic SDK 或 Agent SDK 的开发场景,自动识别项目语言并给出对应示例与默认配置,快速搭建 LLM 应用。
✎ 想把Claude能力接进应用或智能体,用claude-api上手快、兼容Anthropic与Agent SDK,集成路径清晰又省心
提示工程专家
by alirezarezvani
覆盖Prompt优化、Few-shot设计、结构化输出、RAG评测与Agent工作流编排,适合分析token成本、评估LLM输出质量,并搭建可落地的AI智能体系统。
✎ 把提示优化、LLM评测到RAG与智能体设计串成一套方法,适合想系统提升AI开发效率的人。
智能体流程设计
by alirezarezvani
面向生产级多 Agent 编排,梳理顺序、并行、分层、事件驱动、共识五种工作流设计,覆盖 handoff、状态管理、容错重试、上下文预算与成本优化,适合搭建复杂 AI 协作系统。
✎ 帮你把多智能体流程设计、编排和自动化统一起来,复杂工作流也能更稳地落地,适合追求强控制力的团队。
相关 MCP 服务
顺序思维
编辑精选by Anthropic
Sequential Thinking 是让 AI 通过动态思维链解决复杂问题的参考服务器。
✎ 这个服务器展示了如何让 Claude 像人类一样逐步推理,适合开发者学习 MCP 的思维链实现。但注意它只是个参考示例,别指望直接用在生产环境里。
知识图谱记忆
编辑精选by Anthropic
Memory 是一个基于本地知识图谱的持久化记忆系统,让 AI 记住长期上下文。
✎ 帮 AI 和智能体补上“记不住”的短板,用本地知识图谱沉淀长期上下文,连续对话更聪明,数据也更可控。
PraisonAI
编辑精选by mervinpraison
PraisonAI 是一个支持自反思和多 LLM 的低代码 AI 智能体框架。
✎ 如果你需要快速搭建一个能 24/7 运行的 AI 智能体团队来处理复杂任务(比如自动研究或代码生成),PraisonAI 的低代码设计和多平台集成(如 Telegram)让它上手极快。但作为非官方项目,它的生态成熟度可能不如 LangChain 等主流框架,适合愿意尝鲜的开发者。