← 返回 Skills 市场

医生工作台助手 (Agentic AI 科研平台)

Name: 医生工作台助手 (Agentic AI 科研平台)
Author: emergenceronearth

作者 EmergencerOnEarth · GitHub ↗ · v0.1.0 · MIT-0

cross-platform ⚠ suspicious

105

总下载

当前安装

版本数

在 OpenClaw 中安装

/install agentic-doctor-assistant

功能描述

医生工作台助手。整理患者基本信息、关键时间线、检验趋势、待办事项并生成随访任务草稿。当用户以医生身份查询患者情况或需要随访建议时触发。

使用说明 (SKILL.md)

医生工作台助手 Skill

何时使用

当用户以医生角色请求患者相关信息时使用，例如：

「帮我整理这个患者最近的情况」
「查看患者时间线」
「生成随访任务」

本 skill 独立于科研任务流程，可单独触发。

执行步骤

1. 上报开始

curl -s -X POST http://localhost:5001/api/report \
  -H "Content-Type: application/json" \
  -d '{"skill":"doctor-assistant","display_name":"医生工作台助手","status":"running","message":"正在整理患者摘要、时间线与检验趋势..."}'

2. 读取数据并输出

读取 /home/ubuntu/workspace/demo/mock_data/doctor_demo.json，向用户展示：

患者基本信息：姓名（脱敏）、性别、年龄、主诊断、主治医生
关键事件时间线：按时间顺序列出每个事件的日期、类型、摘要
检验指标趋势：列出各日期的 WBC、CRP（如有 PCT 也列出）数值，标注趋势方向
待办事项清单：按优先级列出每条待办
随访任务草稿：
- 任务标题
- 建议随访时间
- 关键随访问题列表

3. 上报完成

curl -s -X POST http://localhost:5001/api/report \
  -H "Content-Type: application/json" \
  -d '{"skill":"doctor-assistant","display_name":"医生工作台助手","status":"completed","message":"患者摘要已整理，随访任务草稿已生成"}'

安全使用建议

Before installing, confirm where the skill will read patient data and where it will send reports. The SKILL.md hardcodes /home/ubuntu/workspace/demo/mock_data/doctor_demo.json and posts to http://localhost:5001 but the skill metadata does not declare these paths/endpoints. Ask the publisher to (1) declare required config paths or environment variables for the data file and reporting endpoint, (2) explain and enforce de-identification and logging/consent controls for PHI, and (3) provide a configurable/reporting endpoint rather than a hardcoded localhost URL. Only install if you trust the local report service, the data source, and if you have organizational approval to let an agent read patient records. If you cannot verify those, treat this skill as risky and do not enable it for real patient data.

功能分析

Type: OpenClaw Skill Name: agentic-doctor-assistant Version: 0.1.0 The skill functions as a medical data summarizer, reading a local JSON file and reporting status updates to a local API endpoint (localhost:5001). Its behavior is entirely consistent with its stated purpose in SKILL.md, and it lacks any indicators of data exfiltration, unauthorized execution, or malicious intent.

能力评估

⚠ Purpose & Capability

The skill claims to summarize patient records and generate follow-up tasks, which legitimately requires reading patient data. However, the SKILL.md hardcodes a specific filesystem path (/home/ubuntu/workspace/demo/mock_data/doctor_demo.json) and localhost reporting endpoint (http://localhost:5001) while the skill metadata declares no required config paths, endpoints, or credentials. The metadata and instructions are not aligned.

⚠ Instruction Scope

Runtime instructions instruct the agent to: 1) POST status messages to localhost:5001, and 2) read a local JSON file containing patient data and present de-identified info. These steps access potentially sensitive PHI and an external (local) endpoint. The SKILL.md does not define how de-identification is performed, how to validate the file, nor any limits on what else may be read; it also lacks guidance about consent, logging, or secure handling.

✓ Install Mechanism

This is an instruction-only skill with no install spec or bundled code, so no files are written or third-party packages installed by the skill itself.

⚠ Credentials

The skill requests no environment variables or credentials, yet its instructions require filesystem access to a specific path and the ability to call a local HTTP service. Those resource requirements are not declared; reading patient data is sensitive and should be justified and scoped via explicit config (e.g., declared config paths or env vars for data location and report endpoint).

✓ Persistence & Privilege

The skill is not always-enabled and does not request elevated persistence. It is user-invocable and can run autonomously per platform defaults — not itself a red flag, but combine with the other concerns about data access.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install agentic-doctor-assistant
安装完成后，直接呼叫该 Skill 的名称或使用 /agentic-doctor-assistant 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v0.1.0

initial publish

元数据

Slug agentic-doctor-assistant

版本 0.1.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题