← Back to Skills Marketplace
tujinsama

自动客服应答虾

by Ricky · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
78
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install auto-customer-service-claw
Description
自动客服应答虾 — 处理客户咨询的自动化回复系统。核心能力:意图识别与分类、知识库匹配、智能转人工判断、多轮对话管理。当以下情况时使用此 Skill:(1) 用户要求搭建或配置自动客服系统,(2) 需要处理客户咨询的自动化回复(售前、售后、账户问题等),(3) 需要配置 FAQ 知识库、意图识别规则或回复话术,(...
README (SKILL.md)

自动客服应答虾

概述

自动处理客户咨询,通过意图识别 + 知识库匹配实现秒级响应,复杂问题智能转人工。

工作流程

步骤 1:接收客户消息

监听客服渠道(飞书、微信、网站在线客服、APP 内客服等),实时接收客户消息。

步骤 2:意图识别与分类

读取 references/intent-rules.md,分析客户消息,识别咨询意图:

  • 售前咨询(商品信息、价格、库存)
  • 售后问题(退换货、物流、发票)
  • 账户问题(登录、密码、会员)
  • 投诉建议(产品问题、服务不满)
  • 闲聊寒暄(打招呼、感谢)

步骤 3:知识库匹配

读取 references/faq-database.md,根据识别的意图搜索最匹配的答案:

  • 精确匹配:关键词完全匹配
  • 语义匹配:使用向量相似度计算
  • 模糊匹配:处理拼写错误、同义词
  • 上下文匹配:结合多轮对话历史

步骤 4:生成回复内容

读取 references/response-templates.md,根据匹配结果生成回复:

  • 标准答案:直接返回知识库中的答案
  • 动态答案:调用 API 获取实时数据(如库存、物流)
  • 引导式回复:通过多轮对话收集信息
  • 转人工提示:无法处理时引导转人工

步骤 5:智能转人工判断

识别需要转人工的场景:

  • 连续 3 次未匹配到答案
  • 客户明确要求人工客服
  • 检测到负面情绪(愤怒、失望)
  • 涉及退款、投诉等敏感问题

步骤 6:数据记录

记录每次对话(客户问题、匹配答案、客户反馈、是否转人工),用于持续优化知识库。

服务管理

使用 scripts/chatbot-server.sh 管理客服机器人服务:

# 启动服务
./scripts/chatbot-server.sh start

# 停止服务
./scripts/chatbot-server.sh stop

# 重新加载知识库
./scripts/chatbot-server.sh reload

# 测试问答效果
./scripts/chatbot-server.sh test "这个商品有货吗?"

知识库管理

使用 scripts/knowledge-updater.sh 更新知识库:

# 从 Excel 导入 FAQ
./scripts/knowledge-updater.sh import --file faq.xlsx

# 导出当前知识库
./scripts/knowledge-updater.sh export --output current_faq.xlsx

自定义配置

  • 改 FAQ 知识库 → 编辑 references/faq-database.md
  • 改意图识别规则 → 修改 references/intent-rules.md
  • 改回复话术 → 修改 references/response-templates.md

环境依赖

  • Python 3.8+
  • pip install flask transformers
  • 飞书插件(如接入飞书客服渠道)
Usage Guidance
This skill is internally coherent, but before running it consider: (1) The embedded Flask server binds to 0.0.0.0 and the /chat endpoint has no authentication — run it behind a firewall, reverse proxy with auth, or bind to localhost if you don't want external access. (2) The knowledge-updater will append data from user-supplied Excel/CSV/MD files directly into the FAQ markdown — only import trusted files to avoid polluting the KB. (3) There are no channel connectors included; integrating Feishu/WeChat requires additional secure credentials and connector code. (4) Run the service in a sandbox/container and restrict network access if you plan to expose it, and review logs in /tmp for sensitive information. If you want stronger guarantees, add authentication to the /chat API and validation/sanitization when importing FAQs.
Capability Analysis
Type: OpenClaw Skill Name: auto-customer-service-claw Version: 1.0.0 The skill bundle implements a functional customer service automation system but contains significant security vulnerabilities. Specifically, `scripts/chatbot-server.sh` and `scripts/knowledge-updater.sh` exhibit shell injection flaws by directly interpolating variables (such as `$SERVER_PORT` and `$file`) into Python code strings executed via `python3 -c` without any sanitization. While these appear to be unintentional coding errors rather than intentional malware, they represent a high-risk surface for arbitrary code execution if inputs are manipulated. The core logic in `SKILL.md` and the reference files remains aligned with the stated purpose.
Capability Assessment
Purpose & Capability
Name/description (自动客服应答) match the provided files: FAQ, intent rules, response templates, a server management script and a knowledge-updater script. The requested artifacts and scripts are proportional to building a lightweight chatbot and knowledge import/export.
Instruction Scope
SKILL.md stays within the stated purpose (intent classification, KB matching, templates, transfer-to-human rules). It mentions integrating channels (Feishu, WeChat, website) but the included scripts do not implement connectors — they provide a local HTTP /chat endpoint and instructions to 'use a Feishu plugin' for channel integration. This is not incoherent but is an implementation gap the operator must fill.
Install Mechanism
No install spec; the skill is instruction+scripts only. Dependencies are standard (Python, Flask, optionally openpyxl). No network downloads or obscure install URLs are included.
Credentials
The skill requests no environment variables or credentials. The only runtime env referenced is optional CHATBOT_PORT — this matches the local server role. No unrelated secrets or config paths are requested.
Persistence & Privilege
always:false and user-invocable default behavior. The skill does not request persistent platform-wide privileges or modify other skills. It writes PID/log files to /tmp and updates its own references files — expected for a local service.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install auto-customer-service-claw
  3. After installation, invoke the skill by name or use /auto-customer-service-claw
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
首次发布:自动客服应答系统,支持意图识别、知识库匹配、智能转人工
Metadata
Slug auto-customer-service-claw
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is 自动客服应答虾?

自动客服应答虾 — 处理客户咨询的自动化回复系统。核心能力:意图识别与分类、知识库匹配、智能转人工判断、多轮对话管理。当以下情况时使用此 Skill:(1) 用户要求搭建或配置自动客服系统,(2) 需要处理客户咨询的自动化回复(售前、售后、账户问题等),(3) 需要配置 FAQ 知识库、意图识别规则或回复话术,(... It is an AI Agent Skill for Claude Code / OpenClaw, with 78 downloads so far.

How do I install 自动客服应答虾?

Run "/install auto-customer-service-claw" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is 自动客服应答虾 free?

Yes, 自动客服应答虾 is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does 自动客服应答虾 support?

自动客服应答虾 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created 自动客服应答虾?

It is built and maintained by Ricky (@tujinsama); the current version is v1.0.0.

💬 Comments