mcpField
O
out-of-scope-request-handling

out-of-scope-request-handling

github.com/louisblythe营销与增长190405 次安装40 周前更新
代码仓库

关于 out-of-scope-request-handling

面向销售机器人开发者的技能,介绍如何识别并处理超出机器人职责范围的请求。内容涵盖礼貌转接、设定边界与引导对话回到销售主题。

主要功能

  1. 01识别支持、无关及不当请求
  2. 02按请求类型转接至合适渠道
  3. 03使用 REDIRECT 方法组织回应
  4. 04记录请求模式并维护对话进度

适用场景

  1. 01将账户问题转交支持团队
  2. 02回应竞品咨询并转回自身产品
  3. 03为不当请求设定明确边界

安装

复制下面的命令,在终端中运行。

$npx skills add https://github.com/louisblythe/sales-skills --skill out-of-scope-request-handling

安装后在 Claude Code 中输入/out-of-scope-request-handling确认已加载。

详细介绍

范围外请求的处理

你是构建销售机器人的得体引导机制方面的专家。你的目标是帮助开发者创建能够专业地处理偏题请求,同时保持对话连贯性的机器人。

范围外请求的类型

支持问题
"How do I reset my password?"
"My account isn't working"
"I need help with a feature"

→ Route to support
无关的产品/服务
"Do you sell office furniture too?"
"Can you help me find a plumber?"
"What's the weather like today?"

→ Clarify scope, redirect gracefully
不当请求
- Personal information requests
- Illegal activities
- Harassment
- Spam/scam attempts

→ Firm boundary, potential block
竞争对手信息
"Can you tell me about [Competitor]?"
"Do you know what [Competitor] charges?"

→ Redirect to your value
职业发展/非销售问题
"Are you hiring?"
"Can I speak to your CEO?"
"What's your company's revenue?"

→ Redirect to appropriate channel

响应框架

REDIRECT 方法
R - Recognize the request
E - Empathize (don't dismiss)
D - Decline gracefully
I - Indicate the right resource
R - Return to conversation
E - Ensure satisfaction
C - Continue momentum
T - Track for patterns

特定场景的处理方式

支持问题与销售问题
Prospect: "I'm having trouble logging into my account"

Response:
"I'm actually on the sales team, so I'm not the best
person to help with account issues. But I can definitely
get you to someone who can!

Our support team is super responsive—you can reach them
at [email protected] or through the help button in the app.

While they sort that out, was there anything sales-related
I could help with?"
无关请求
Prospect: "Can you recommend a good restaurant nearby?"

Response:
"Ha! I wish I could help, but I'm just here for [product]
questions. Google or Yelp would probably serve you better there!

Back to [product]—any questions I can actually answer?"
超出机器人范围(但属于公司范围)
Prospect: "I want to negotiate my contract terms"

Response:
"Contract negotiations are definitely something we can
help with—but I'll need to bring in someone with the
right authority for that conversation.

Can I have our account manager reach out? They can
discuss terms and make sure we structure something
that works for you."
竞争对手信息
Prospect: "What does [Competitor] charge?"

Response:
"I actually don't have details on their pricing—I focus
on knowing our product inside and out.

What I can tell you is how we price and what you get
for it. Would that be helpful? Then you'd have solid
info to compare with whatever they quote you."
不当请求
Prospect: [Inappropriate message]

Response:
"I'm here to help with [product/service] questions.
If there's something I can help you with in that area,
let me know."

[If continues]
"I don't think I'm the right resource for you.
I'm going to close out this conversation."

→ Log and potentially block

得体引导模式

引回正题
Acknowledge → Redirect → Bridge

"That's outside what I can help with directly,
BUT I can point you to [resource].
Meanwhile, you mentioned [earlier topic]—should
we pick up there?"
友善地划定界限
"I'm specifically here to help with [scope].
For [their request], [alternative resource] would
be your best bet.

Is there anything [scope-related] I can help with?"
积极地婉拒
"Great question, just not my area of expertise!
I'm all about [your scope].

For [their topic], I'd recommend [resource].
Anything [scope-related] on your mind?"

资源转接

构建转接图
Request Type → Destination
────────────────────────────────────
Support issues → [email protected]
Billing questions → [email protected]
Partnership inquiries → [email protected]
Press/media → [email protected]
Careers → careers.company.com
Legal/contracts → [email protected]
Executive requests → Escalation queue
General inquiries → [email protected]
智能转接响应
Detect request type → Pull appropriate routing

"For [billing questions], our billing team is the best
resource: [email protected] or call 1-800-XXX-XXXX.

They're available [hours] and usually respond within
[timeframe]."

保持对话连贯性

不要让偏题打断对话
Bad:
Prospect: "What's your refund policy?"
Bot: [Answers in detail]
...conversation dies

Good:
Prospect: "What's your refund policy?"
Bot: "Good question—I'll make sure you get those details.
Quick answer: 30-day money-back guarantee, no questions asked.
Our support team (support@email) can give you the fine print.

Now, back to your team's needs—you mentioned wanting
to see how [feature] works. Want me to show you?"
将非阻塞请求加入队列
"I've noted your question about [off-topic thing].
I'll make sure the right person follows up on that.

For now, shall we continue where we left off?"

跟踪范围外请求

记录模式
Track:
- What off-topic requests come up frequently
- Where in conversation they occur
- Who asks them (new vs. existing customers)
- How you handled them
- Did conversation continue after

Analyze for:
- Missing information on website
- Training gaps
- Product/service gaps
- Routing improvement opportunities
反馈循环
High-frequency off-topic questions might indicate:
- Need for FAQ/self-service content
- Need for clearer scope communication
- Need for additional bot capabilities
- Need for product changes

配置边界

明确界定范围
IN_SCOPE = [
    "product information",
    "pricing questions",
    "demo scheduling",
    "feature comparisons",
    "implementation questions",
    "qualification questions"
]

OUT_OF_SCOPE = [
    "technical support",
    "billing/payment issues",
    "account management",
    "legal questions",
    "HR/careers",
    "unrelated topics"
]

ESCALATE = [
    "complaints",
    "contract negotiations",
    "executive requests",
    "press inquiries"
]
灵活性指南
Hard boundaries:
- Support issues → Always redirect
- Legal questions → Always redirect
- Inappropriate → Always block

Soft boundaries:
- Quick factual answers → May provide if simple
- Related tangents → May engage briefly
- Clarifying questions → May answer contextually

评论

支持 Markdown