Veterinary Clinic Bundle

VetClaw - 52个宠物医院/兽医诊所AI自动化技能套装。覆盖排班管理、病历记录、客户回访、经营分析等场景。

Security Scan

The bundle's code mostly matches a veterinary assistant purpose, but the package metadata omits multiple environment credentials and there are runtime instructions that will send data to third‑party APIs (LLM, SMS, WeChat) — a mismatch that warrants caution.

Purpose & Capability

The skill name/description (vet clinic automation) aligns with the included code and handlers (appointment, intake, emergency, inventory, etc.). However, the registry metadata declares no required env vars/credentials while the configuration files and code clearly expect multiple external service credentials (DEEPSEEK_API_KEY, SMS_API_KEY/SMS_API_SECRET, WECHAT_APP_ID/WECHAT_APP_SECRET, DB_PASSWORD, etc.). That omission is inconsistent and reduces transparency.

Instruction Scope

SKILL.md and README provide sensible setup steps (edit config/vet-config.yaml, set LLM provider, run installer). They instruct storing data locally or on your cloud. But runtime instructions and code cause the agent/server to read configuration and environment variables, persist PII/medical records into a local SQLite DB, and — if configured — transmit user data to external services (DeepSeek LLM, Aliyun SMS, WeChat, payment APIs). The docs do not explicitly warn that medical/personal data may be sent to third‑party LLMs or SMS/WeChat providers when those integrations are enabled.

Install Mechanism

There is no external installer that downloads code from unknown URLs; the bundle is code-included and contains a Python install script and a FastAPI app. No remote archive/extract/install URLs are present in the manifest. Installing means running included Python scripts and pip installing listed Python dependencies (requirements.txt).

Credentials

Although the registry lists no required env vars, the code and config reference multiple secret-bearing environment variables (DEEPSEEK_API_KEY, SMS_API_KEY/SMS_API_SECRET, WECHAT_APP_ID/WECHAT_APP_SECRET, DB_PASSWORD). Those are legitimate for the described integrations, but the metadata omission is a transparency problem. Providing these keys will cause user text (potentially PII or sensitive medical symptom descriptions) to be transmitted to third‑party services; that risk should be explicit. The number of secret/credential types is moderate but must be disclosed.

Persistence & Privilege

The skill does not request forced 'always' installation and does not attempt to modify other skills. It creates a local SQLite DB (./data/vetclaw.db) and writes skill files via the included installer script — normal for a local web app. There is no code attempting to alter other agent configs or escalate privileges.

What to consider before installing

This bundle appears to implement the described vet-clinic features, but be aware of these issues before installing: - Missing declared credentials: The package metadata says no env vars required, but the code/config expect several secrets (DEEPSEEK_API_KEY, SMS_API_KEY/SMS_API_SECRET, WECHAT_APP_ID/WECHAT_APP_SECRET, optional DB_PASSWORD). Ask the publisher to update the registry to list required env vars before providing keys. - Third‑party data flow: If you set an LLM provider API key (DeepSeek/Qwen/OpenAI) or enable SMS/WeChat, user messages and medical symptom details will be sent to those external services. Only use API keys for providers you trust and ensure that sending clinical or personal data to them complies with your privacy/legal requirements. - Run in an isolated/test environment first: The bundle creates a local SQLite DB and starts a web service. Test on a non-production host and inspect logs/traffic to ensure no unexpected outbound endpoints. - Review code for integration endpoints: Confirm the external endpoints (e.g., https://api.deepseek.com) are legitimate and acceptable for your data policy. If you need to avoid external LLMs, run with no API key so the code falls back to local/template behavior. - Ask for provenance: Source/homepage is 'unknown' but README references a GitHub repo and support email; verify the project repository and author identity before trusting production data. What would raise confidence: explicit registry declaration of required env vars, a published source repository (with commit history and maintainer info), and a privacy statement clarifying what data is sent to which third parties. If you want, I can: (1) list every point in code where an outbound network call is made, (2) extract all environment variable names the code reads, or (3) suggest minimal config settings to run the app offline (no external APIs).

VetClaw — 宠物医院AI技能套装 SKILL.md

概述

VetClaw 是一套专为宠物医院/诊所设计的52个AI自动化技能，基于 OpenClaw 平台运行。

快速配置

1. 诊所基础信息

编辑 config/vet-config.yaml：

clinic:
  name: "你的宠物医院名称"
  address: "详细地址"
  phone: "联系电话"
  hours:
    weekday: "09:00-21:00"
    weekend: "10:00-18:00"
  services:
    - type: "基础诊疗"
      price_range: "50-200"
    - type: "疫苗接种"
      price_range: "80-300"
    - type: "绝育手术"
      price_range: "500-2000"
    - type: "急诊"
      price_range: "200-5000"

2. LLM 配置

llm:
  provider: "deepseek"  # 可选: deepseek, qwen, openai
  api_key: "${DEEPSEEK_API_KEY}"
  model: "deepseek-chat"
  temperature: 0.3

3. 通知渠道

notifications:
  sms:
    provider: "aliyun"  # 阿里云短信
    api_key: "${SMS_API_KEY}"
    sign_name: "你的宠物医院"
  wechat:
    app_id: "${WECHAT_APP_ID}"
    app_secret: "${WECHAT_APP_SECRET}"

技能详细说明

📞 前台接待类

vet-intake — 新客户信息采集

• 触发词: "新客户"、"登记"、"建档"
• 功能: 通过对话采集客户姓名、联系方式、宠物信息（品种/年龄/性别/是否绝育）
• 输出: JSON格式客户档案，自动存入数据库

vet-appointment — 预约挂号

• 触发词: "预约"、"挂号"、"看诊"
• 功能: 根据排班表自动推荐时段，处理预约冲突，发送确认通知
• 排班冲突检测: 自动检查医生排班，避免超额预约

vet-phone-ai — AI电话接听

• 触发词: 来电自动触发
• 功能: 识别来电意图（预约/咨询/价格/投诉），自动应答或转人工
• 话术模板: 内置50+宠物医院常见问题应答模板

vet-emergency — 急诊分流

• 触发词: "急"、"危险"、"出血"、"抽搐"、"呕吐不止"
• 功能: 根据症状描述判断紧急程度（红/黄/绿三级），指导客户处理
• 注意: 严重情况自动转人工，AI不做诊断

📋 病历与诊疗类

vet-lab-interpret — 化验单AI解读

• 触发词: "化验单"、"血常规"、"生化"
• 功能: 输入化验数值，AI标注异常指标并用通俗语言解读
• 支持: 血常规、生化全套、尿检、便检

vet-vaccine-schedule — 疫苗计划

• 触发词: "打疫苗"、"免疫计划"
• 功能: 根据宠物种类/年龄生成疫苗接种时间表，自动设置提醒

💰 收费与财务类

vet-daily-report — 每日营收报表

• 触发词: "今日营收"、"日报"
• 功能: 自动汇总当日收费项目、总金额、各支付方式占比
• 输出: 结构化报表 + 可视化图表

📢 营销与客户类

vet-churn-predict — 流失预警

• 触发词: "流失客户"、"久未复诊"
• 功能: 分析客户就诊间隔，标记超过平均间隔1.5倍的客户
• 动作: 自动生成关怀话术，建议联系

知识库

兽医知识库 (vet-knowledge-base.json)

{
  "common_diseases": {
    "犬瘟热": {
      "symptoms": ["发热", "咳嗽", "流鼻涕", "腹泻"],
      "urgency": "high",
      "advice": "立即就医，隔离其他宠物"
    }
  },
  "vaccine_schedules": {
    "dog": {
      "幼犬": ["6-8周: 第一针联苗", "10-12周: 第二针联苗", "14-16周: 第三针联苗+狂犬"],
      "成年犬": ["每年: 联苗加强+狂犬"]
    }
  },
  "drug_interactions": {
    "阿莫西林": {
      "avoid_with": ["氨基糖苷类抗生素"],
      "notes": "与食物同服可减少胃肠反应"
    }
  }
}

工作流集成

典型工作流：新客户到院 → 就诊 → 回访

客户来电/到店
  ↓
vet-phone-ai (判断意图)
  ↓
vet-intake (采集信息)
  ↓
vet-appointment (安排预约)
  ↓
vet-reminder (预约前1天提醒)
  ↓
[到院就诊]
  ↓
vet-record-create (创建病历)
  ↓
vet-prescription (开处方)
  ↓
vet-invoice (生成费用)
  ↓
vet-discharge (出院指导)
  ↓
vet-followup (设置复诊提醒)
  ↓
vet-feedback (满意度调查)

与其他技能包的兼容性

• ai-secretary-bundle: 共享日程/提醒模块
• content-creator-bundle: 社交媒体内容可直接使用
• ecommerce-bundle: 支付/订单模块可复用

常见问题

Q: AI会代替兽医做诊断吗？

A: 不会。VetClaw 只处理行政和沟通任务，诊断和治疗决策始终由执业兽医完成。

Q: 数据安全吗？

A: 所有数据存储在本地或您选择的云服务，不会共享给第三方。符合《个人信息保护法》要求。

Q: 需要什么技术基础？

A: 基本的电脑操作能力即可。安装脚本会自动完成配置。

支持

• 文档: https://github.com/DMNO1/vetclaw-bundle/wiki
• 问题反馈: https://github.com/DMNO1/vetclaw-bundle/issues
• 技术支持: [e-mail]