home / skills / aaaaqwq / agi-super-skills / healthcare-monitor

healthcare-monitor skill

/skills/healthcare-monitor

This skill monitors healthcare company financing signals in real-time, identifies investment signals, and pushes alerts to Telegram and Feishu.

npx playbooks add skill aaaaqwq/agi-super-skills --skill healthcare-monitor

Review the files below or copy the command above to add this skill to your agents.

Files (40)
SKILL.md
7.6 KB
---
name: healthcare-monitor
description: 医疗行业企业融资监控系统。实时监控医疗健康企业的工商变更,识别融资信号,自动推送告警。支持天眼查/企查查数据采集、AI融资判断、多渠道推送。
allowed-tools: Bash, Read, Write, Edit, Exec, Browser, Message, Cron
---

# 医疗行业企业融资监控系统

## 概述

本技能实现医疗健康行业企业的融资信号实时监控,通过工商变更数据识别未披露的融资事件,为媒体、投资机构提供情报服务。

## 核心功能

### 1. 企业监控列表管理
- 添加/删除监控企业
- 按行业、地区、规模分类
- 设置监控优先级

### 2. 工商变更检测
- 定时爬取天眼查/企查查
- 增量对比检测变更
- 识别关键变更类型

### 3. 融资信号识别
- AI 判断是否为融资
- 融资轮次推断
- 投资方识别

### 4. 多渠道推送
- Telegram 实时告警
- 飞书群组推送
- 日报/周报生成

## 文件结构

```
skills/healthcare-monitor/
├── SKILL.md                 # 技能说明
├── config/
│   ├── companies.json       # 监控企业列表
│   └── settings.json        # 配置参数
├── scripts/
│   ├── monitor.py           # 主监控脚本
│   ├── scraper.py           # 数据采集
│   ├── analyzer.py          # 融资分析
│   └── notifier.py          # 推送通知
├── data/
│   ├── snapshots/           # 企业快照
│   ├── changes/             # 变更记录
│   └── reports/             # 生成的报告
└── templates/
    ├── alert.md             # 告警模板
    └── daily_report.md      # 日报模板
```

## 快速开始

### 1. 添加监控企业

```bash
# 添加单个企业
python3 ~/clawd/skills/healthcare-monitor/scripts/monitor.py add "北京某某医疗科技有限公司"

# 批量导入
python3 ~/clawd/skills/healthcare-monitor/scripts/monitor.py import companies.csv
```

### 2. 手动检查

```bash
# 检查所有企业
python3 ~/clawd/skills/healthcare-monitor/scripts/monitor.py check

# 检查单个企业
python3 ~/clawd/skills/healthcare-monitor/scripts/monitor.py check "公司名称"
```

### 3. 设置定时任务

```bash
# 每小时检查重点企业
# 使用 OpenClaw cron 工具设置
```

## 融资信号识别规则

### 强信号 (置信度 >80%)

| 信号 | 描述 | 权重 |
|------|------|------|
| 新增机构股东 | 股东名称包含"投资/资本/基金/创投" | +40 |
| 注册资本增加 | 增幅 >10% | +30 |
| 股权稀释 | 创始人持股比例下降 | +20 |

### 中信号 (置信度 50-80%)

| 信号 | 描述 | 权重 |
|------|------|------|
| 新增自然人股东 | 可能是投资人代持 | +15 |
| 经营范围扩大 | 业务扩张信号 | +10 |
| 办公地址变更 | 搬迁到更好区域 | +10 |

### 弱信号 (置信度 <50%)

| 信号 | 描述 | 权重 |
|------|------|------|
| 仅注册资本增加 | 可能是内部增资 | +10 |
| 法人变更 | 可能是内部调整 | +5 |

### 融资轮次判断

```yaml
种子/天使轮:
  - 公司成立 <2 年
  - 注册资本 <500 万
  - 新股东为天使投资人

A轮:
  - 公司成立 2-4 年
  - 注册资本增幅 20-50%
  - 新股东为知名 VC

B轮+:
  - 公司成立 >3 年
  - 注册资本增幅 10-30%
  - 新股东为大型 PE/战略投资者
```

## 数据采集

### 天眼查采集

```python
# 采集字段
fields = {
    "basic": ["公司名称", "法人", "注册资本", "成立日期", "经营状态"],
    "shareholders": ["股东名称", "持股比例", "认缴出资", "出资日期"],
    "changes": ["变更时间", "变更项目", "变更前", "变更后"],
    "investments": ["被投企业", "投资金额", "投资时间"],
}
```

### 反爬策略

```yaml
策略:
  - 使用 Playwright 无头浏览器
  - 随机延迟 3-10 秒
  - 轮换 User-Agent
  - 代理 IP 池 (可选)
  - 模拟人类行为 (滚动、点击)

限制:
  - 每小时最多 100 次请求
  - 单企业每天最多查询 3 次
  - 遇到验证码暂停并告警
```

## 推送配置

### Telegram 推送

```json
{
  "telegram": {
    "enabled": true,
    "bot_token": "从 config 读取",
    "chat_id": "8518085684",
    "alert_level": "high"
  }
}
```

### 飞书推送

```json
{
  "feishu": {
    "enabled": true,
    "webhook": "从 config 读取",
    "alert_level": "all"
  }
}
```

## 告警模板

### 融资告警

```markdown
🚨 **融资信号告警**

**企业**: {{company_name}}
**变更时间**: {{change_date}}
**置信度**: {{confidence}}%

**变更详情**:
- 注册资本: {{old_capital}} → {{new_capital}}
- 新增股东: {{new_shareholders}}
- 股权变化: {{equity_changes}}

**AI 分析**:
- 融资轮次: {{round_estimate}}
- 融资金额: {{amount_estimate}}
- 投资方: {{investors}}

**数据来源**: 天眼查
**监控时间**: {{monitor_time}}
```

## API 接口

### 查询企业状态

```bash
GET /api/company/{company_name}

Response:
{
  "name": "公司名称",
  "last_check": "2026-02-03 22:00:00",
  "status": "normal",
  "recent_changes": [...],
  "funding_signals": [...]
}
```

### 添加监控

```bash
POST /api/monitor
{
  "company_name": "公司名称",
  "priority": "high",
  "check_interval": "hourly"
}
```

## 使用场景

### 场景一: 媒体情报

动脉网等医疗媒体使用,第一时间获取融资新闻线索。

```yaml
工作流:
  1. 系统检测到融资信号
  2. 推送到编辑 Telegram
  3. 编辑核实后发布新闻
  4. 比竞争对手快 24-48 小时
```

### 场景二: 投资机构

VC/PE 使用,发现潜在投资标的或跟投机会。

```yaml
工作流:
  1. 系统检测到 A 轮融资
  2. 推送到投资经理
  3. 评估是否跟进
  4. 主动联系企业
```

### 场景三: 大药企 BD

药企商务拓展部门使用,寻找合作/收购标的。

```yaml
工作流:
  1. 系统检测到目标领域企业融资
  2. 推送到 BD 团队
  3. 评估合作可能性
  4. 发起商务接触
```

## 监控企业列表 (初始)

### 医疗器械

- 迈瑞医疗
- 联影医疗
- 微创医疗
- 先瑞达医疗
- 心脉医疗

### 创新药

- 百济神州
- 信达生物
- 君实生物
- 再鼎医药
- 和黄医药

### 医疗AI

- 推想科技
- 数坤科技
- 深睿医疗
- 汇医慧影
- 医渡云

### 互联网医疗

- 微医集团
- 丁香园
- 春雨医生
- 好大夫在线
- 平安好医生

### 基因检测

- 华大基因
- 贝瑞基因
- 燃石医学
- 泛生子
- 诺禾致源

## 配置参数

```json
{
  "check_interval": {
    "high_priority": "1h",
    "normal": "6h",
    "low": "24h"
  },
  "alert_threshold": {
    "confidence": 60,
    "capital_change_percent": 10
  },
  "scraper": {
    "delay_min": 3,
    "delay_max": 10,
    "max_requests_per_hour": 100
  },
  "notification": {
    "telegram": true,
    "feishu": true,
    "email": false
  }
}
```

## 故障排查

### 问题: 爬取失败

```bash
# 检查日志
tail -f ~/clawd/skills/healthcare-monitor/data/logs/scraper.log

# 常见原因
- IP 被封: 更换代理
- 验证码: 人工处理或接入打码平台
- 页面结构变化: 更新选择器
```

### 问题: 误报过多

```bash
# 调整置信度阈值
# 编辑 config/settings.json
{
  "alert_threshold": {
    "confidence": 70  # 提高阈值
  }
}
```

## 相关文件

- `~/clawd/docs/solutions/healthcare-data-intelligence.md` - 完整解决方案
- `~/.openclaw/agents/healthcare-monitor/` - Agent 配置
- `~/clawd/skills/healthcare-monitor/data/` - 数据存储

## TODO

- [ ] 实现天眼查爬虫
- [ ] 实现融资信号 AI 分析
- [ ] 设置定时任务
- [ ] 对接飞书推送
- [ ] 添加 Web 管理界面
- [ ] 支持更多数据源 (企查查、启信宝)
- [ ] 实现日报/周报自动生成

Overview

This skill monitors financing signals for healthcare and medical companies by tracking public corporate registry changes and sending real-time alerts. It identifies likely financing events with an AI-based scoring model and pushes notifications across channels like Telegram and Feishu. The system supports scheduled scraping, incremental change detection, and automated report generation.

How this skill works

The skill periodically scrapes business registry data (e.g., Tianyancha, Qichacha) and stores snapshots for incremental diffing to detect changes in shareholders, registered capital, legal representative, and address. An AI analyzer assigns signal weights and estimates financing likelihood and probable round based on change patterns and company metadata. When confidence exceeds configured thresholds, notifier modules format alerts and push them to configured channels, and daily/weekly reports are produced.

When to use it

  • Monitor portfolio or target lists for early financing signals
  • Provide reporters with pre-vetted leads for funding news
  • Detect M&A or partnership triggers from corporate changes
  • Supply BD or corporate development teams with potential collaboration targets
  • Automate surveillance for compliance or competitive intelligence

Best practices

  • Maintain a prioritized company list and set shorter check intervals for high-priority targets
  • Tune confidence and capital-change thresholds to reduce false positives
  • Respect scraping rate limits and anti-bot measures; use headless browsers and proxy pools responsibly
  • Keep data snapshots and change logs for auditability and retrospective analysis
  • Configure multiple notification channels and escalation rules to avoid missed alerts

Example use cases

  • A healthcare journalist receives a Telegram alert about a sudden capital increase and publishes a scoop before competitors
  • A VC scout finds an unannounced A-round indicator and initiates outreach to the target company
  • A BD team tracks financing in a therapeutic area to identify timely partnership opportunities
  • A compliance team monitors portfolio companies for ownership or legal-representative changes that impact risk profiles

FAQ

What data sources does it support?

It is designed to ingest public corporate registry platforms such as Tianyancha and Qichacha, with a pluggable scraper interface for other sources.

How are financing rounds inferred?

The AI combines rule-based heuristics (age, capital change, investor type) with weighted signals from detected changes to estimate likely round and confidence.

How to reduce false positives?

Raise the confidence threshold, require multiple strong signals before alerting, and refine signal weights in the settings file.