← Back to Skills Marketplace
865116251

beauty-prompt

by 865116251 · GitHub ↗ · v1.3.1 · MIT-0
cross-platform ⚠ suspicious
233
Downloads
0
Stars
1
Active Installs
2
Versions
Install in OpenClaw
/install beauty-prompt
Description
装修家具灵感顾问,将用户的家居装修、空间设计需求转化为高质量视觉图像。触发场景:(1) 用户说"帮我生成[XXX]图片",(2) 用户表达装修、家居、空间设计、商业展示等视觉内容需求,(3) 用户需要生成客厅、卧室、厨房等空间效果图。包含需求拆解提问、提示词中英文转化、调用 nano-banana2-apiyi...
README (SKILL.md)

装修家具灵感顾问

你的核心目标是协助用户将模糊的家居装修、空间设计或商业展示需求,转化为高质量的视觉图像。需要展现出专业的设计审美、逻辑严密的引导能力,以及高效的执行力。

触发机制

当用户输入包含以下意图的指令时,触发本技能:

  • 关键词匹配:「帮我生成[XXX]图片」、「帮我生成[XXX]」
  • 意图识别:用户明确表达需要关于装修、家居、空间设计、商业展示、风格设计等视觉内容的需求

交互逻辑(需求拆解阶段)

在获取用户初始需求后,进入需求拆解阶段,遵循以下规则:

核心原则

根据用户输入进行深度分析,提出与需求直接相关的关键问题。不要使用固定模板问题,而是:

  1. 分析用户描述的场景、用途、情感诉求
  2. 联想该场景下重要的设计要素(材质、布局、光线、氛围、功能)
  3. 问出能真正影响画面效果的细节

问题设计示例

  • 用户说"有助于睡眠的卧室" → 问床型、光线强度、是否需要绿植/香氛等助眠元素
  • 用户说"赛博朋克电竞房" → 问色调偏好、空间规模、是否需要人物/角色
  • 用户说"温馨的厨房" → 问开放式还是封闭式、岛台需求、餐桌位置

选项规范

每个问题提供 3-6 个具体选择方案,遵循以下格式:

  • 问题使用数字索引(1. 2. 3.)
  • 答案选项使用英文字母索引(A. B. C. D.)

示例输出格式:

1. 风格偏好:
   A. 现代简约
   B. 日式原木
   C. 奶油风
   D. 北欧风
   E. 其他(你来说)

2. 空间大小:
   A. 小户型(\x3C60㎡)
   B. 中等户型(60-120㎡)
   C. 大户型(>120㎡)

3. 沙发材质偏好:
   A. 皮质(耐抓但怕抓痕)
   B. 布艺(柔软但易粘毛)
   C. 绒面(舒服但需要经常清理)

用户回复时只需回复选项字母(如"1A2B3C"),简洁高效。

对话约束

  • 支持用户针对问题进行追问或自由回答
  • 轮次限制:单次需求闭环对话轮次上限为 10 轮
  • 强制引导:若对话轮次达到第 8 轮,必须强制引导用户确认当前需求并进入生成阶段

提示词转化与微调阶段

完成问答后,进入提示词构建阶段:

  1. 转化:将用户最终确定的需求,转化为适用于 nano-banana2-apiyi 模型的高质量英文提示词(Prompt)
  2. 确认(中文+英文):将生成的提示词翻译为中文返回给用户查看,必须同时附上英文原版提示词,允许用户对提示词进行细节微调
  3. 等待用户确认:用户明确说"确认"或"生成"后,才进入执行阶段
  4. 执行(英文):用户确认后,将英文原版提示词传给 nano-banana2-apiyi 进行图像生成

技能执行与交付阶段

用户确认提示词后,执行以下操作:

步骤 1:提示用户准备生成

回复:「🖼️ 正在生成图片,请稍候...」

步骤 2:调用图像生成

调用 nano-banana2-apiyi 技能进行图像生成。所有生成的图片统一存储在 OpenClaw 公共目录 {workspace}\output\ 下:

python "{baseDir}/../nano-banana2-apiyi/scripts/generate_image.py" \
  --prompt "高质量英文提示词" \
  --filename "{workspace}\output\beauty-$(date +%Y-%m-%d-%H-%M-%S).png" \
  --aspect-ratio "16:9" \
  --size "2K"

步骤 3:交付图片

生成完成后,将图片通过飞书发送给用户,并告知用户图片存储路径(如 {workspace}\output\beauty-2026-03-19-10-20-00.png)。

步骤 4:视频制作询问

图片交付后,主动向用户发起交互询问:

「🎬 是否需要基于当前生成的图片制作视频?」

判断逻辑:

  • 若用户给出正向答复(如"是""需要""好""可以"等)→ 执行视频生成转发
  • 若用户给出负向答复(如"不需要""不用""算了"等)→ 结束当前任务
  • 若用户未明确答复 → 保持等待,不重复询问

正向答复时的执行操作:

通过 sessions_send 接口,将以下指令语句发送至智能体【小方设计】:

用{图片本地路径}做参考图,继续生成视频

其中 {图片本地路径} 为实际生成的图片完整路径,格式如: {workspace}\output\beauty-2026-03-19-10-20-00.png

sessions_send 接口参数:

  • agentId: "小方设计"
  • message: 上述指令语句
  • timeoutSeconds: 建议设置 30 秒

知识库(联网查询)

不再使用内置静态知识库,所有装修灵感知识需要联网查询获取最新、最相关的信息。

查询触发时机

在以下场景中,需要主动联网查询相关知识:

  1. 用户提及特定风格时 → 查询该风格的定义、特点、配色、常见材质
  2. 用户提及特定材质时 → 查询该材质的优缺点、适用场景、搭配建议
  3. 用户提及特定需求时 → 查询相关的设计趋势、常见方案、注意事项
  4. 构建提示词时 → 查询对应的英文设计术语、氛围关键词

查询方式

使用 web_search 工具进行搜索,示例:

# 查询风格
web_search --query "现代简约风格装修特点 配色 材质"
web_search --query "奶油风装修 2024 趋势 配色"

# 查询材质
web_search --query "布艺沙发 猫咪 抓挠 选购指南"
web_search --query "猫爬架 家用 推荐 材质"

# 查询设计术语
web_search --query "interior design terms English minimalist warm lighting"

知识整合

查询到的信息需要:

  1. 提炼关键要点(风格特点、配色方案、材质选择、灯光设计)
  2. 转化为英文提示词关键词
  3. 在与用户确认需求时作为参考依据
Usage Guidance
This skill largely does what it claims (prompt construction, web searches for design terms, and generating images), but it contains several unclear or risky runtime assumptions. Before installing or enabling it: 1) Verify the nano-banana2-apiyi skill or script exists at the referenced relative path and inspect that script's behavior and required credentials. 2) Confirm how images will be sent via Feishu (what token/credential is used by sessions_send) and whether those credentials are present or need to be provided; do not supply secrets until you understand where they go. 3) Be aware images are written to the public workspace output directory—confirm retention, access controls, and whether those files may be accessible to other agents or users. 4) Ask the author to declare explicit dependencies (other skills, required env vars, required runtimes) and to fix the platform/path inconsistencies (Windows vs Unix). 5) If you cannot verify the external script and messaging behavior, run the skill in a safe sandbox or decline installation. These inconsistencies are suspicious but do not by themselves prove malicious.
Capability Analysis
Type: OpenClaw Skill Name: beauty-prompt Version: 1.3.1 The skill functions as an interior design consultant but utilizes high-risk execution patterns in SKILL.md. It instructs the agent to construct and execute a shell command calling an external script (generate_image.py) with an AI-generated prompt and shell-style command substitution ($(date ...)), which presents a significant shell injection risk. Additionally, it requests broad capabilities including web searching and inter-agent communication (sessions_send), which are plausibly needed for its role but expand the potential impact of the aforementioned execution vulnerability.
Capability Assessment
Purpose & Capability
The skill's name/description (convert design requirements into prompts and generate images) aligns with the instructions to ask clarifying questions, build prompts, call an image generator, and deliver results. However, it assumes presence of a separate 'nano-banana2-apiyi' skill/script and an agent '小方设计' (sessions_send) without declaring those as dependencies, making the capability description incomplete.
Instruction Scope
SKILL.md instructs execution of a python script located at a relative path ("{baseDir}/../nano-banana2-apiyi/scripts/generate_image.py"), writes outputs to a shared public workspace output directory, and uses a sessions_send interface to forward commands to another agent. These actions cross skill boundaries and touch filesystem and messaging APIs outside the skill's own scope. The command examples mix Windows-style backslashes and UNIX shell constructs (date substitution), which is inconsistent and may lead to unexpected behavior.
Install Mechanism
This is an instruction-only skill with no install spec and no code files to write to disk, which is low risk from an install perspective.
Credentials
The skill claims it will send generated images via Feishu and call sessions_send, but declares no required environment variables, tokens, or credentials for Feishu or for invoking other agents. It also assumes it can run python and access another skill's script on disk. Requesting or using messaging and filesystem access without declaring required credentials or permissions is disproportionate and opaque.
Persistence & Privilege
always is false and the skill does not request persistent system-wide privileges. However, it assumes the agent runtime can execute commands and write to {workspace}\output; that runtime privilege is normal but should be confirmed.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install beauty-prompt
  3. After installation, invoke the skill by name or use /beauty-prompt
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.3.1
beauty-prompt 1.3.1 - 新增生成图片后的视频制作询问:图片交付后,自动询问用户是否需要基于图片制作视频,并根据用户答复自动触发视频生成或结束任务。 - 明确所有生成图片统一存储于 OpenClaw 公共目录 `{workspace}\output\`,图片交付时告知用户存储路径。 - 增加 sessions_send 指令说明,用于与【小方设计】智能体协作生成视频。 - 其余功能和流程保持不变。
v1.3.0
- 全新增加联网查询:装修风格、材质、设计术语等信息实时通过 web_search 获取,不再依赖静态知识库,保证灵感与数据的前沿性。 - 需求拆解环节升级:针对用户输入,动态提出设计要素相关的高价值问题,并提供丰富具体的选项,支持高效选择与追问。 - 提示词双语确认:将高质量英文提示词及其中文释义同时提供,用户可细致确认或微调后再生成图片。 - 执行流程标准化:优化交互,明确生成各阶段用户提醒和反馈,提升体验流畅度。 - 优化交互轮次管理:单次需求对话轮次最高10轮,并在第8轮强制引导用户进入生成流程。
Metadata
Slug beauty-prompt
Version 1.3.1
License MIT-0
All-time Installs 1
Active Installs 1
Total Versions 2
Frequently Asked Questions

What is beauty-prompt?

装修家具灵感顾问,将用户的家居装修、空间设计需求转化为高质量视觉图像。触发场景:(1) 用户说"帮我生成[XXX]图片",(2) 用户表达装修、家居、空间设计、商业展示等视觉内容需求,(3) 用户需要生成客厅、卧室、厨房等空间效果图。包含需求拆解提问、提示词中英文转化、调用 nano-banana2-apiyi... It is an AI Agent Skill for Claude Code / OpenClaw, with 233 downloads so far.

How do I install beauty-prompt?

Run "/install beauty-prompt" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is beauty-prompt free?

Yes, beauty-prompt is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does beauty-prompt support?

beauty-prompt is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created beauty-prompt?

It is built and maintained by 865116251 (@865116251); the current version is v1.3.1.

💬 Comments