Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Med-LLM-Wiki

v1.0.0

管理医疗领域患者临床咨询知识库,结构化存储患者信息、问答、干预方案及循证概念,支持精准细节检索与推理。

0· 90·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for zhangyingf/med-llm-wiki.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Med-LLM-Wiki" (zhangyingf/med-llm-wiki) from ClawHub.
Skill page: https://clawhub.ai/zhangyingf/med-llm-wiki
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install med-llm-wiki

ClawHub CLI

Package manager switcher

npx clawhub@latest install med-llm-wiki
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
high confidence
Purpose & Capability
Name/description (medical patient consultation knowledge base) match the SKILL.md instructions to parse documents, create patient/consultation/concept pages, build indexes, and answer queries. However the SKILL.md assumes read/write access to a local wiki/ directory but the registry metadata declares no required config paths; this mismatch is worth noting (the skill implicitly requires filesystem access even though nothing is declared).
!
Instruction Scope
The instructions explicitly direct the agent to ingest raw patient-specific clinical documents, preserve all numeric values and specific recommendations, create multiple files under wiki/, and use those files as sources when answering queries. There are no instructions about de-identification, consent, encryption, access control, retention policy, or logging. For a skill that handles protected health information (PHI), the absence of any data-protection guidance is a material concern.
Install Mechanism
This is an instruction-only skill with no install spec and no code files — lowest install risk. Nothing is downloaded or written by an installer step according to the registry metadata.
Credentials
The skill requests no environment variables, credentials, or binaries, which is proportional to a purely file-based wiki manager. However, because it handles sensitive medical data, the absence of any instruction to use secure storage credentials or a dedicated secrets store (or to document where files will be written) is noteworthy — it may lead to insecure file placement or accidental exposure.
Persistence & Privilege
always:false (normal). The skill allows model invocation (default), which combined with its instructions means it could autonomously read/write the wiki/ files when invoked. There is no request to modify other skills or system-wide settings. Still, autonomous access to PHI heightens the operational risk, so review invocation policies and runtime environment.
What to consider before installing
This skill is coherent with its stated goal of creating and querying a structured medical wiki, but it explicitly instructs the agent to ingest and persist patient-identifiable clinical data without any privacy or security guidance. Before installing or enabling it, you should: 1) ensure the agent runs in a secure, access-controlled environment (HIPAA-compliant if applicable); 2) define exactly where the wiki/ files will be stored and who can access them (avoid writing to shared or cloud-exposed paths); 3) add or require mechanisms for de-identification or explicit patient consent if you will store PHI; 4) consider adding encryption-at-rest, access logging, and retention/deletion policies; 5) restrict autonomous invocation or require user confirmation before creating or returning any patient-specific content; and 6) if you need stricter control, request the author add explicit config options for secure storage and auditability. If you cannot guarantee secure handling of sensitive health data in your environment, do not enable this skill.

Like a lobster shell, security has layers — review code before you run it.

latestvk971cecfpc99zt34m59y75s2js858g5k
90downloads
0stars
1versions
Updated 1w ago
v1.0.0
MIT-0

name: "Med-llm-wiki" description: "管理和维护医疗领域的 LLM Wiki 知识库,支持高精度细节检索与循证推理,针对真实患者病历、临床咨询文档、医患问答进行深度结构化。"

Med LLM Wiki Manager

本技能用于将以患者为中心的临床咨询文档(包含患者背景、多主题问答、方案建议)转化为高度互联的 Wiki 知识网络。

核心设计:四层知识模型

采用以下分层策略:

层级Wiki 页面类型对应内容检索命中目标
L1患者实体页患者人口学、既往史、基线检查值“张三的 FEV1 是多少?”
L2临床问答页每个独立的“问题+回答”作为一个页面“关于呼吸训练有哪些建议?”
L3干预方案页回答中具体的药物/训练/时间表“术前蛋白补充的具体剂量?”
L4概念/证据页疾病指南、药物说明书、循证依据“GOLD 2 级 COPD 围术期管理原则?”

关键原则问题与回答永不合并。每个问题都值得独立成页,以保留完整的上下文和推理链条。

目录结构

wiki/
├── index.md                  # 全局索引(按患者/问题类别/关键数值索引)
├── log.md                    # 操作日志
├── patients/                 # L1: 患者实体页
│   └── patient-xxx.md
├── encounters/               # 就诊/检查事件页
│   └── encounter-xxx.md
├── consultations/            # L2: 【核心新增】临床问答页
│   ├── consult-xxx-q01.md    # 每个问题独立文件
│   └── consult-xxx-q02.md
├── interventions/            # L3: 具体干预方案页(可选,复杂方案独立成页)
│   └── intervention-xxx-protein.md
├── concepts/                 # L4: 背景知识概念页
│   └── concept-copd-gold.md
└── comparisons/              # 对比分析页

页面类型详细规范

1. 患者实体页 (patients/)

用途:锚定所有信息的核心对象。仅包含静态背景信息基线检查数据摘要

示例结构 (patients/patient-72male-vats.md):

---
title: "患者:72岁男性 VATS 右肺上叶切除术候选"
patient_id: p20240415-001
age: 72
gender: male
diagnosis:
  - "右肺占位 (拟行 VATS 右肺上叶切除术)"
  - "COPD (GOLD 2级)"
  - "高血压病"
  - "轻度低蛋白血症"
  - "中度焦虑状态 (GAD-7=10)"
baseline_data:
  fev1: "65% 预计值"
  dlco: "60%"
  6mwt: "320 m"
  albumin: "32 g/L"
  d_dimer: "1.8 mg/L"
  hr: "88 bpm (窦性,偶发房早)"
  bp: "138/86 mmHg"
  spo2_rest: "94%"
  spo2_activity: "88%"
  sleep_awakenings: "3-4 次/夜"
consultation_questions:
  - "[[consult-p20240415-q01]]"
  - "[[consult-p20240415-q02]]"
  # ... 共10个
created: 2026-04-20
updated: 2026-04-20
---

# 患者摘要:72岁男性 VATS 术前评估

## 背景
- 吸烟史:40年,已戒5年。
- 手术计划:VATS 右肺上叶切除术。

## 基线检查结果
(略,详见 Frontmatter 或内嵌表格)

2. 临床问答页 (consultations/) —— 核心原子单元

用途:存储一个完整的“问题 + 回答”对。这是 LLM 回答细节问题的第一手素材

必须包含的元数据

  • patient_id: 关联患者。
  • question_id: 原始文档中的问题编号。
  • question_text: 问题的原文。
  • keywords: 用于索引的关键词(如:呼吸训练、低蛋白、焦虑)。
  • recommendation_strength: 方案推荐强度/证据等级(若原文有)。

示例结构 (consultations/consult-p20240415-q02.md):

---
title: "咨询:如何通过呼吸训练改善FEV1值应对术后肺复张?"
type: consultation
patient_id: p20240415-001
question_id: 2
question_text: "如何通过呼吸训练改善FEV1值应对术后肺复张?"
keywords: [呼吸训练, FEV1, 肺复张, 术前康复, COPD]
related_concepts: ["[[concept-incentive-spirometry]]", "[[concept-prehab-copd]]"]
sources: ["病例1.md"]
created: 2026-04-20
---

# 咨询:呼吸训练改善 FEV1 与术后肺复张

## 问题原文
如何通过呼吸训练改善FEV1值应对术后肺复张?

## 回答摘要
为改善 FEV1 和术后肺复张,需综合呼吸训练、营养优化、心理干预及术前风险评估。

## 核心推荐方案

### 1. 术前呼吸训练组合
- **缩唇呼吸**:吸气与呼气时间比 1:2。
- **腹式呼吸**:每日 2 次,每次 15 分钟。
- **激励式肺量计训练**:目标容积 ≥15 ml/kg,每小时 10 次。
- **阈值负荷训练**:强度为最大吸气压的 30-50%。

### 2. 围术期优化要点
- **氧疗支持**:维持 SpO₂ ≥92%,活动时可短暂提高至 95%。
- **营养补充**:高蛋白 + 维生素 D + 抗氧化剂。
- **心理管理**:渐进式肌肉放松训练。

### 3. 术后肺复张策略
- **早期下床活动**:术后 6 小时床旁坐立,24 小时内下床。
- **体位引流**:右侧卧位结合前倾坐位咳嗽。
- **持续气道正压 (CPAP)**:预防肺不张。

## 关键执行要点
- **启动时间**:术前 7 天启动强化训练。
- **多学科协作**:呼吸治疗师、物理治疗师共同参与。

## 风险预警
- D-二聚体升高者训练中需监测血栓风险。
- 房早患者避免 Valsalva 动作。

> 来源:病例1.md - 问题2

3. 干预方案页 (interventions/) —— 可选高级抽象

用途:当某个回答中的方案非常具体且可复用(如“术前蛋白质补充策略”),可单独建页,便于未来其他患者引用。

示例 (interventions/intervention-preop-protein-supplement.md):

---
title: "术前低蛋白血症营养干预方案"
type: intervention
target_condition: "低蛋白血症 (白蛋白 <35 g/L)"
applicable_population: "择期大手术患者"
protocol:
  target_albumin: "≥35 g/L"
  protein_intake: "1.5-2.0 g/kg/d"
  supplementation: "乳清蛋白 40-50 g/d + 酪蛋白 20 g 睡前"
  monitoring: "每3天检测前白蛋白"
evidence_level: "基于临床指南及研究"
---
(具体步骤略)

4. 概念页 (concepts/)

用途:解释背景知识,如疾病分级、检查意义、药物机制。不包含患者特异性数据

示例 (concepts/concept-copd-gold.md):

---
title: "COPD GOLD 分级与围术期风险"
type: concept
related_guidelines: ["GOLD 2024"]
---

# COPD GOLD 分级

## GOLD 2 级 (中度)
- FEV1: 50% ≤ FEV1 < 80% 预计值。
- 围术期风险:肺部并发症风险增加,需术前优化。
(略)

摄入新文件:分析存储患者咨询

当摄入此类患者背景 + 多问题咨询文档时,严格遵循以下步骤:

Step 1: 解析与分割

  1. 识别患者背景区块(从“我今年72岁…”到“睡眠片段化”)。
  2. 识别问题列表(如“1,吸烟戒断5年患者是否需要特殊呼吸道准备?”)。
  3. 将每个问题的回答内容完整切分,保留所有子标题、数值和引用

Step 2: 创建/更新页面

  1. 患者实体页 (patients/):从背景区块提取静态数据写入 Frontmatter 和正文。
  2. 临床问答页 (consultations/):为每一个问题创建一个独立文件。
    • 使用模板填充 Frontmatter。
    • 将回答内容原样整理进正文,可适当添加小标题以提升可读性,但严禁删减数值和具体建议
  3. 概念页 (concepts/):若回答中反复出现特定术语(如 GOLD 分级、DLCO),则创建或更新对应概念页。
  4. 干预方案页 (interventions/):若某回答的步骤性极强(如蛋白质补充时间表),可选择独立建页并建立链接。

Step 3: 建立交叉引用

  • 患者页 → 链接所有问答页。
  • 问答页 → 链接相关概念页和干预方案页。
  • 问答页之间若有逻辑顺序(如问题 1 和问题 2 相关),在正文中注明 参见 [[consult-xxx-q01]]

Step 4: 更新索引

index.md 中:

  • 按患者索引:列出该患者所有问答页。
  • 按主题索引:列出所有问答页的关键词标签云。
  • 数值索引:将关键基线数值(FEV1、白蛋白)与患者页关联。

知识库查询:细节问题精准定位

当用户发起提问后,严格遵循以下步骤:

检索路径

  1. 读取 index.md → 定位关键词 → 找到相关问答页。
  2. 读取相关问答页
  3. 若问答页链接了 interventions/ 页,可进一步读取获取结构化方案。
  4. 生成回答,直接引用原文数值并注明来源页面。

回答示例

根据病例记录,针对该72岁患者的术前低蛋白血症(白蛋白 32 g/L),建议的蛋白质补充策略为:每日总蛋白摄入 1.5-2.0 g/kg(若体重70kg则需105-140 g/d),其中乳清蛋白粉每日40-50 g 分两次冲服。目标是在术前7-10天将白蛋白提升至 ≥35 g/L。详见 [[consult-p20240415-q03]]。

知识库维护

  • 每个问答页是否完整保留了原始回答中的所有数值和单位
  • 患者页的 Frontmatter 中是否包含了所有关键检查结果?
  • 问答页之间是否存在未链接的依赖关系?(如问题6的答案引用了问题2的呼吸训练,需建立双向链接)
  • 是否存在重复的概念解释,需统一抽象到概念页?
  • 索引中的关键词是否覆盖了用户可能提问的所有角度?

附录:完整页面结构示例

摄入新文件后,预期生成以下 Wiki 文件:

wiki/
├── index.md
├── patients/
│   └── patient-72male-vats.md
├── consultations/
│   ├── consult-p20240415-q01-respiratory-prep.md
│   ├── consult-p20240415-q02-breath-training.md
│   ├── consult-p20240415-q03-protein-supplement.md
│   ├── consult-p20240415-q04-anxiety-nonpharma.md
│   ├── consult-p20240415-q05-early-ambulation-oxygen.md
│   ├── consult-p20240415-q06-anticoagulation-d-dimer.md
│   ├── consult-p20240415-q07-muscle-training-analgesia.md
│   ├── consult-p20240415-q08-anxiety-pain-perception.md
│   ├── consult-p20240415-q09-diaphragm-exercise.md
│   └── consult-p20240415-q10-postural-drainage.md
├── interventions/
│   ├── intervention-preop-protein.md
│   └── intervention-diaphragm-training-post-lobectomy.md
├── concepts/
│   ├── concept-copd-gold.md
│   ├── concept-dlco.md
│   ├── concept-gad7.md
│   └── concept-enhanced-recovery-after-surgery.md
└── log.md

Comments

Loading comments...