← Back to Skills Marketplace
unisound-llm

unisound-pe-missing-negative

by Unisound-LLM · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
57
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install unisound-pe-missing-negative
Description
门诊病历内涵质控:体格检查缺少与诊断相关的体征。给定门诊病历文本,调用内部医疗大模型,输出无缺陷或有缺陷及原因。
README (SKILL.md)

体格检查缺少与诊断相关的体征

概述

本 skill 仅针对**「体格检查缺少与诊断相关的体征」**这一条内涵质控规则。将字段解析、多步 LLM 推理与条件分支实现在本目录 scripts/emr_qc_impl.py 中;scripts/emr_qc.pyscripts/run.py 为入口,本 skill 可单独打包发布。调用 HiVoice MaaS 医疗大模型(OpenAI 兼容 chat/completions),输出 无缺陷有缺陷 + 原因。

数据安全、隐私与伦理声明

  • 最小必要原则:仅处理质控所必需的病历文本内容;不要求提供姓名、证件号等身份信息。
  • 严格脱敏:在发送至任何接口前,请确保病历文本已完成脱敏/去标识化处理。
  • 不做本地持久化:不将输入与中间结果写入本地持久化存储;本次调用结束即销毁
  • 医疗边界:本技能为辅助质控工具,不构成医疗诊断或治疗建议;最终结论请由执业医生审核。

输入格式

纯文本门诊病历(UTF-8),各字段用中文标签开头,例如:

主诉:发热3天
现病史:患者3天前无明显诱因出现发热,体温最高38.5℃...
既往史:高血压病史5年,规律服用氨氯地平,血压控制可...
体格检查:T 38.2℃,BP 150/90mmHg,双肺呼吸音清...
初步诊断:上呼吸道感染
处理意见:对症治疗

支持的字段标签:主诉现病史既往史体格检查辅助检查初步诊断/诊断处理意见/处理(中英文冒号均可)。

也支持通过 scripts/run.py 直接输入 pdf/doc/docx/xls/xlsx/csv/txt/json

快速开始

# 文本入口(在仓库 skills 根目录运行)
python3 doctor/emr-qc/pe-missing-negative/scripts/emr_qc.py \
  --input data/med-emr-qc/record.txt \
  --appkey \x3Cyour-appkey>

# 多格式入口
python3 doctor/emr-qc/pe-missing-negative/scripts/run.py \
  --input data/med-emr-qc/record.xlsx \
  --appkey \x3Cyour-appkey>

参数说明

scripts/emr_qc.py

  • --input PATH必填。门诊病历文本文件路径(UTF-8)。
  • --appkey STRING必填。调用内部医疗大模型的鉴权 key,由平台分配;不得写入仓库
  • --base URL:大模型 base URL(默认:https://maas-api.hivoice.cn/v1)。
  • --model STRING:模型名称(默认:u1-insuremed)。
  • --timeout SECONDS:HTTP 超时秒数;0 表示一直等待(默认:0)。
  • --output PATH:输出文件路径(默认:../runs/med-emr-qc/pe-missing-negative.txt)。

scripts/run.py 附加参数:

  • --input-type auto|pdf|doc|docx|xls|xlsx|csv|txt|json:输入类型;默认 auto
  • --sheet STRING:读取 Excel 时指定 sheet(可选)。
  • --encoding STRINGtxt/csv 编码(默认:utf-8)。
  • --save-prepared:保存预处理后的文本,便于调试。

输出约定

  • 输出为 UTF-8 文本:
    • 无缺陷:输出 无缺陷
    • 有缺陷:输出 有缺陷 + 换行 + 原因说明

依赖

自包含实现

质控与 LLM 调用逻辑均在 scripts/emr_qc_impl.py(与同 skill 一并发布)。

预处理(仅 run.py

scripts/run.py 依赖 _shared/doc-preprocesspreprocess.py),须位于仓库 skills/ 根下。

运行环境

  • Python 3.7+
  • 无需第三方包(仅标准库;多格式输入时可选 openpyxlpypdf

外部 API

  • 医疗大模型:https://maas-api.hivoice.cn/v1/chat/completions(POST,OpenAI 兼容格式)
Usage Guidance
Install only if you are comfortable sending de-identified physical-exam and diagnosis text to the configured HiVoice MaaS endpoint. Protect the app key, avoid the debug save option for real patient data, and verify the shared preprocessing helper before using non-text file inputs.
Capability Analysis
Type: OpenClaw Skill Name: unisound-pe-missing-negative Version: 1.0.0 The skill bundle is a medical record quality control tool designed to identify missing physical examination findings relative to diagnoses. The implementation in `emr_qc_impl.py` and `emr_qc.py` uses standard Python libraries to parse text and interact with a specific medical LLM API (maas-api.hivoice.cn). The code and instructions in `SKILL.md` are consistent with the stated purpose, include data privacy warnings, and show no evidence of malicious intent, data exfiltration, or unauthorized execution.
Capability Assessment
Purpose & Capability
The purpose, code, and documentation are coherent: the skill reads a user-provided outpatient record, checks a specific physical-exam quality rule, and calls a disclosed medical LLM. The notable risk is that medical content is sent to an external provider.
Instruction Scope
Instructions emphasize de-identification and medical-review boundaries. There is a minor privacy caveat because the documentation says inputs are not persisted, while an explicit debug option can save preprocessed text.
Install Mechanism
There is no install script or automatic package execution. The multi-format entry point imports a shared preprocessor outside this skill's manifest, so that helper should be trusted if the multi-format path is used.
Credentials
Local file reads, output file writes, and network access to the LLM are proportionate to the stated workflow. Users should keep inputs scoped and de-identified.
Persistence & Privilege
The skill does not request OS privileges or background persistence, but it requires a service app key and writes the QC result; the optional --save-prepared flag stores the preprocessed input text.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install unisound-pe-missing-negative
  3. After installation, invoke the skill by name or use /unisound-pe-missing-negative
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release: provides quality control for outpatient EMRs, focusing on missing physical exam findings relevant to the diagnosis. - Text and multi-format input support, including PDF, DOC/DOCX, XLS/XLSX, CSV, TXT, JSON. - Uses internal HiVoice MaaS medical models via OpenAI-compatible API. - Includes strong data privacy controls: strict de-identification, minimal necessary processing, and no local storage. - Outputs “无缺陷” (no defect) or “有缺陷” (defect present) with reasons. - Command-line interface with flexible parameters for file input, model selection, and output.
Metadata
Slug unisound-pe-missing-negative
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is unisound-pe-missing-negative?

门诊病历内涵质控:体格检查缺少与诊断相关的体征。给定门诊病历文本,调用内部医疗大模型,输出无缺陷或有缺陷及原因。 It is an AI Agent Skill for Claude Code / OpenClaw, with 57 downloads so far.

How do I install unisound-pe-missing-negative?

Run "/install unisound-pe-missing-negative" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is unisound-pe-missing-negative free?

Yes, unisound-pe-missing-negative is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does unisound-pe-missing-negative support?

unisound-pe-missing-negative is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created unisound-pe-missing-negative?

It is built and maintained by Unisound-LLM (@unisound-llm); the current version is v1.0.0.

💬 Comments