← Back to Skills Marketplace
aaiccee

chronic-disease-review

by aaiccee · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
241
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install chronic-disease-review
Description
门诊慢病审核(糖尿病/高血压)。输入 OCR 结果数组 JSON,输出审核结论与原因(原始 JSON + 自然语言结论)。
README (SKILL.md)

门诊慢病审核

概述

给定一份 OCR 结果数组(每项包含 fileName/page/docType/ocrText),本技能会:

  • 输出审核接口原始 JSON
  • 输出自然语言摘要(结论 + 原因)

数据安全、隐私与伦理声明

  • 最小必要原则:仅处理审核所必需的文本内容;不要求也不鼓励提供与审核无关的身份信息。
  • 严格脱敏:在发送至任何模型/接口前,会对可识别个人身份的信息进行脱敏/去标识化处理(如姓名、证件号、手机号、详细地址、人脸/影像等)。仅传递脱敏后的必要信息用于本次 skill 调用。
  • 不做本地持久化:不将用户输入与中间结果写入本地持久化存储(包含磁盘文件、数据库、日志)。仅在内存中短暂处理;本次调用结束即销毁
  • 第三方 API 风险提示:在功能需要时,可能会调用第三方模型/服务接口;此时仅会发送脱敏后的必要信息,并使用加密传输。除完成本次请求外,不用于任何其他用途(如训练、画像、营销)。
  • 医疗边界:本技能输出为审核规则匹配与原因摘要的辅助信息,不构成医疗诊断或治疗建议;如涉及临床判断请以执业医生意见为准。

输入格式

输入必须是 JSON 数组(list),示例:

[
  {"fileName":"xxx.pdf","page":1,"docType":"出院记录","ocrText":"..."},
  {"fileName":"xxx.pdf","page":2,"docType":"检验报告","ocrText":"..."}
]

快速开始

skills 目录执行:

# 糖尿病
python3 scripts/chronic_disease_review.py \
  --disease-code diabetes \
  --input data/med-chronic-disease-review/糖尿病_ocr_task_result.json

# 高血压
python3 scripts/chronic_disease_review.py \
  --disease-code hypertension \
  --input data/med-chronic-disease-review/高血压_ocr_task_result.json

参数说明

  • --disease-code STRING必填。糖尿病/高血压(也支持 diabetes/hypertension/dm/htn 别名)。本 skill 每次只审核一个病种。
  • --review-type STRING:可选。默认 慢病审核(也可传 大病审核 等服务支持的值)。
  • --input PATH:OCR 数组 JSON(UTF-8)。
  • --base URL:后端 base(默认:http://10.10.20.15:9011)。
  • --llm-model STRING:可选。模型名。
  • --timeout SECONDS:HTTP 超时;0 表示一直等待(默认:0)。
  • --output-json PATH:保存原始返回 JSON(默认:../runs/med-chronic-disease-review/{scenario}_resp.json)。
  • --output-text PATH:保存自然语言摘要(默认:../runs/med-chronic-disease-review/{scenario}_resp.txt)。

输出约定

  • 若输出路径父目录不存在,会自动创建。
  • 自然语言摘要重点字段:
    • final_decision:通过/不通过
    • reasoning:原因(可选)

备注

  • 发布约束:示例输入、运行输出、自测脚本均放在 skill 包外(分别位于 ../data/../runs/../self_tests/),skill 目录内仅保留可发布的核心文件(scripts/SKILL.md_meta.json)。
Usage Guidance
Do not assume the privacy promises in the documentation are enforced. Before using with any real PHI: (1) review and modify the DEFAULT_BASE or always pass --base to point to a trusted internal endpoint; (2) add/verify de-identification of ocrText (remove names, IDs, phone numbers, addresses) before any network calls; (3) if you must avoid local persistence, remove or change the code that writes ../runs/ files or run in a controlled ephemeral environment; (4) audit the remote endpoint's trustworthiness and data handling policy if you plan to use the provided default; (5) consider running the script on synthetic or redacted test data first. The mismatch between SKILL.md and code (default base URL and persistence/de-id assertions) is a red flag — treat this skill as untrusted until corrected and re-audited.
Capability Analysis
Type: OpenClaw Skill Name: chronic-disease-review Version: 1.0.0 The skill exhibits a significant discrepancy between its documentation and implementation regarding data privacy. While SKILL.md claims 'strict desensitization' of PII before transmission, scripts/chronic_disease_review.py sends the raw OCR input directly to an external endpoint (shangbao.yunzhisheng.cn). Additionally, the script writes output files to paths outside its own directory structure (../runs/), which is a risky behavior. These issues represent a high privacy risk and potential data leak, though they appear to be critical vulnerabilities or poor implementation rather than intentional malware.
Capability Assessment
Purpose & Capability
Name/description (chronic disease review of OCR JSON) align with the code: the scripts validate OCR JSON, POST it to a review API, and format the returned review. However the code's DEFAULT_BASE points to an external host (https://shangbao.yunzhisheng.cn/...) while SKILL.md documents a different default base (http://10.10.20.15:9011), which is an unexplained discrepancy and deserves caution.
Instruction Scope
SKILL.md asserts strict de-identification of personally identifiable information before any external calls and promises no local persistence; the code does not implement any de-identification of ocrText and does POST the OCR content verbatim to the remote /api/v1/review/flow/by-ocr. The code also writes raw response JSON and formatted text files to disk (../runs/…), contradicting the 'no local persistence' claim.
Install Mechanism
Instruction-only with two small Python scripts, no install spec, no third-party packages or downloads — low install risk.
Credentials
The skill requests no credentials or env vars, which is appropriate; however it transmits potentially sensitive OCR data to a hard-coded external domain by default. Absence of credential requirements does not mitigate the privacy risk of sending protected health information to an external service without demonstrated de-identification.
Persistence & Privilege
SKILL.md asserts 'no local persistence', but the script explicitly writes raw response JSON and a natural-language summary to disk under ../runs/med-chronic-disease-review. This local persistence is inconsistent with the documented privacy guarantee.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install chronic-disease-review
  3. After installation, invoke the skill by name or use /chronic-disease-review
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
med-chronic-disease-review 1.0.0 - Initial Release - Provides automated outpatient chronic disease (diabetes/hypertension) review based on OCR result arrays. - Outputs both original review JSON and a natural language summary (decision + reasoning). - Ensures strict data privacy: only essential, de-identified data processed transiently in memory. - CLI supports customizable input, disease type selection, model, and output file paths. - Designed as an auxiliary tool for chronic disease audits; not a substitute for clinical judgment.
Metadata
Slug chronic-disease-review
Version 1.0.0
License MIT-0
All-time Installs 1
Active Installs 1
Total Versions 1
Frequently Asked Questions

What is chronic-disease-review?

门诊慢病审核(糖尿病/高血压)。输入 OCR 结果数组 JSON,输出审核结论与原因(原始 JSON + 自然语言结论)。 It is an AI Agent Skill for Claude Code / OpenClaw, with 241 downloads so far.

How do I install chronic-disease-review?

Run "/install chronic-disease-review" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is chronic-disease-review free?

Yes, chronic-disease-review is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does chronic-disease-review support?

chronic-disease-review is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created chronic-disease-review?

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

💬 Comments