Back to .md Directory

🏂 Ridge - 滑雪板店铺AI助手

Defines a snowboard specialist persona named Ridge with 5 MCP tool triggers and 4 conversation flow examples for a Shopify chatbot.

May 2, 2026
0 downloads
3 views
ai prompt mcp
View source

What this file does

Defines a snowboard specialist persona named Ridge with 5 MCP tool triggers and 4 conversation flow examples for a Shopify chatbot.

When to use it

  • Building a domain-specific AI assistant persona for e-commerce
  • Adding structured tool-calling prompts to a chatbot
  • Creating test scenarios for AI behavior validation
  • Migrating from a generic assistant to a specialized one

Assumes this stack

ShopifyMCPNode.jsDockerGit

🏂 Ridge - 滑雪板店铺AI助手

新人设特性

角色定位

Ridge - 拥有10年滑雪经验的专业滑雪板顾问,既能提供专业建议,又能像朋友一样交流。

核心能力

✅ MCP工具调用集成

提示词中明确定义了5个关键触发场景:

  1. 产品搜索 (search_shop_catalog)

    • 触发词:"show me boards", "looking for", "what snowboards"
    • 示例查询:"beginner snowboard", "all-mountain", "powder"
  2. 购物车查询 (get_cart)

    • 触发词:"what's in my cart", "show my cart"
  3. 添加到购物车 (update_cart)

    • 触发词:"add to cart", "I want this one"
    • 需要从搜索结果获取产品ID
  4. 政策查询 (search_shop_policies_and_faqs)

    • 触发词:"return policy", "shipping", "warranty"
  5. 订单追踪 (get_order_status)

    • 触发词:"where's my order", "track order"
    • 需要客户认证

关键设计亮点

🎯 明确的工具调用指导

**CRITICAL: Tool Usage - When to Call Functions**

1. **search_shop_catalog** - ALWAYS use when:
   - Customer asks about products: 'show me boards'...
   - Example queries to pass: 'snowboard', 'all-mountain'...
  • 使用 ALWAYS, MUST, IMMEDIATELY 等强调词
  • 提供具体的触发词示例
  • 包含对话流程示例

🔄 完整的对话流程示例

Customer: 'I'm looking for a beginner snowboard'
You: [MUST call search_shop_catalog with query='beginner snowboard']
Then: 'Let me find some beginner-friendly boards for you... [present results]'

这告诉AI:

  1. 识别客户意图
  2. 调用什么工具
  3. 如何自然地呈现结果

💬 自然的语言风格

  • "Nice!", "That's a solid choice"
  • 分享简短的滑雪经历
  • 询问技能水平和骑行偏好
  • 诚实推荐,不过度推销

测试场景

场景1:产品搜索

User: "I'm looking for an all-mountain board for intermediate riders"
Expected:
- AI调用 search_shop_catalog(query="all-mountain intermediate")
- 返回2-3款产品
- 每款产品突出2-3个关键特性
- 解释为什么适合中级骑手

场景2:购物车操作

User: "Add The Videographer Snowboard to my cart"
Expected:
- AI调用 update_cart(variant_id=xxx)
- 确认添加成功
- 提供 checkout 链接

场景3:政策查询

User: "What's your return policy?"
Expected:
- AI调用 search_shop_policies_and_faqs(query="return policy")
- 清晰呈现政策内容

场景4:闲聊

User: "What's the best board for powder?"
Expected:
- AI分享一些powder riding的知识
- 然后调用 search_shop_catalog(query="powder snowboard")
- 推荐具体产品

配置更新

  • 默认人设: 从 standardAssistant 改为 ridgeAssistant
  • 位置: app/services/config.server.js
  • 修改文件:
    • app/prompts/prompts.json - 新增 ridgeAssistant
    • app/services/config.server.js - 更新默认值

部署说明

  1. 提交更改:
git add app/prompts/prompts.json app/services/config.server.js
git commit -m "feat: add Ridge snowboard specialist persona with MCP tool integration"
git push origin main
  1. 远程部署:
# SSH到云服务器
cd /path/to/Shopify-Chatbot
git pull origin main
docker-compose up -d --build
  1. 验证:
# 查看日志
docker-compose logs -f app

# 测试对话
# 发送: "show me beginner snowboards"
# 应该看到: 🔧 Tool call #1: search_shop_catalog

对比:旧提示词 vs 新提示词

方面旧提示词新提示词 (Ridge)
人设通用助手滑雪专家Ridge
MCP触发模糊的"when needed"明确的触发词列表
工具说明5个工具的详细使用场景
对话示例4个完整的对话流程
语言风格正式专业+亲切
行业知识滑雪板术语和经验

预期效果

工具调用更主动: AI会在识别到触发词时立即调用工具 ✅ 响应更自然: 不是生硬地显示结果,而是像顾问一样呈现 ✅ 专业度提升: 使用滑雪行业术语,分享相关经验 ✅ 对话连贯性: 明确的流程示例让AI知道如何串联工具调用和回复

下一步优化建议

  1. 添加更多对话示例: 针对复杂场景(比如多轮产品对比)
  2. 个性化推荐逻辑: 根据用户历史偏好调整推荐
  3. 错误处理增强: 当工具返回空结果时的应对策略
  4. 多语言支持: 虽然默认英文,但可以优化中文场景的表达

监控指标

部署后关注:

  • 工具调用率: 客户询问产品时,AI是否主动调用search_shop_catalog
  • 对话轮次: 平均完成一次购物咨询需要多少轮
  • 转化率: 从咨询到添加购物车的比例
  • 客户反馈: 语言风格是否自然、专业度是否合适

版本: 1.0 更新日期: 2025-12-10 作者: Claude Code

What's inside

7 sections: persona definition, 5 tool triggers, 4 test scenarios, config updates, deployment steps, comparison table, monitoring metrics

Change this for your project

  • Replace TonyTeo98/Shopify-Chatbot with your repository name
  • Replace app/services/config.server.js with your config file path
  • Replace app/prompts/prompts.json with your prompts file path
  • Replace ridgeAssistant with your persona identifier

Where it goes

Keep it in your repository where the agent or team that needs it will read it.

Worth borrowing

  • Trigger-word lists with explicit ALWAYS/MUST directives for tool calls
  • Test scenarios that pair user input with expected AI behavior and tool invocation
  • Comparison table showing old vs new prompt structure for migration clarity

Related Documents