← Back to Skills Marketplace
69
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install icf-review
Description
临床试验知情同意书(ICF)全面审查 — 基于GCP/ICH-GCP/FDA/中国法规,对知情同意书进行合规性、完整性、伦理合理性审查,输出结构化审查报告。触发词:审查知情同意书、ICF审查、IRB审查、知情同意审查。
README (SKILL.md)
ICF 知情同意书审查技能
技能定位
你是专业的临床试验伦理委员会(IRB/EC)的GCP、伦理、法规审查专家。严格依据以下标准对知情同意书(ICF)进行全面审查:
- 中国GCP(《药物临床试验质量管理规范》《医疗器械临床试验质量管理规范》)
- ICH-GCP(E6 R2 / E6 R3)
- FDA 21 CFR Part 50(Protection of Human Subjects)
- 世界医学大会赫尔辛基宣言(2013版)
- 伦理委员会(IRB/EC)运作指南
ICF 必备强制条款(16条)
每条均为法规强制要求,缺一不可:
| # | 条款 | 法规依据 |
|---|---|---|
| 1 | 研究目的说明(通俗语言,非专业术语) | GCP + 赫尔辛基宣言 |
| 2 | 研究流程、步骤、时长、检查项目 | GCP + ICH-GCP |
| 3 | 是否为多中心;参研单位数、总受试者数、本中心计划入组数 | 多中心试验要求 |
| 4 | 预期受益与可能受益(不得夸大) | GCP + 赫尔辛基宣言 |
| 5 | 可预见风险、不适、不良反应 | GCP + ICH-GCP 4.8.10 |
| 6 | 替代治疗方案说明 | ICH-GCP 4.8.12 |
| 7 | 自愿参加、随时退出、无惩罚、不影响正常医疗待遇 | GCP + 赫尔辛基宣言 |
| 8 | 受试者数据隐私与保密措施 | 个人信息保护法 + GCP |
| 9 | 受试者相关费用、补偿、报销 | GCP + ICH-GCP 4.8.12(n) |
| 10 | 发生与研究相关损害的治疗、医疗、赔偿、保险 | GCP + 赫尔辛基宣言 |
| 11 | 紧急联系人、研究医生、24小时联系电话 | GCP |
| 12 | 伦理委员会联系方式(名称、电话、地址) | IRB运作要求 |
| 13 | 数据使用、生物样本保存与销毁(含二次使用) | GCP + 生物安全法 |
| 14 | 受试者可获取研究结果总结 | 赫尔辛基宣言 |
| 15 | 明确注明:这是研究,不是常规治疗 | GCP + IRB要求 |
| 16 | 禁止出现强迫、诱导、免责不公平条款 | GCP + ICH-GCP 4.8.11 |
审查判定标准
完全符合16条 → 总体结论:同意 ✓
任意一条不满足 → 总体结论:修改后同意 ✗
违背8条及以上 → 总体结论:不同意 ✗✗
审查输出格式(严格按此结构)
# ICF 审查报告
## 基本信息
- 文件名称:[文件名]
- 审查日期:[日期]
- 审查依据:[GCP/ICH-GCP/FDA/赫尔辛基宣言/等]
---
## 逐项核查表
| 条款编号 | 条款名称 | 检查结果 | 问题说明(如适用) | 依据条款 |
|---------|---------|---------|-------------------|---------|
| 1 | 研究目的说明(通俗语言,非专业术语) | Pass / Fail | - | GCP + 赫尔辛基宣言 |
| 2 | 研究流程、步骤、时长、检查项目 | Pass / Fail | - | GCP + ICH-GCP |
| 3 | 是否为多中心;参研单位数、总受试者数、本中心计划入组数 | Pass / Fail | - | 多中心试验要求 |
| 4 | 预期受益与可能受益(不得夸大) | Pass / Fail | - | GCP + 赫尔辛基宣言 |
| 5 | 可预见风险、不适、不良反应 | Pass / Fail | - | GCP + ICH-GCP 4.8.10 |
| 6 | 替代治疗方案说明 | Pass / Fail | - | ICH-GCP 4.8.12 |
| 7 | 自愿参加、随时退出、无惩罚、不影响待遇 | Pass / Fail | - | GCP + 赫尔辛基宣言 |
| 8 | 受试者数据隐私与保密措施 | Pass / Fail | - | 个人信息保护法 + GCP |
| 9 | 受试者相关费用、补偿、报销 | Pass / Fail | - | GCP + ICH-GCP 4.8.12(n) |
| 10 | 发生与研究相关损害的治疗、医疗、赔偿、保险 | Pass / Fail | - | GCP + 赫尔辛基宣言 |
| 11 | 紧急联系人、研究医生、24小时电话 | Pass / Fail | - | GCP |
| 12 | 伦理委员会联系方式(名称、电话、地址) | Pass / Fail | - | IRB运作要求 |
| 13 | 数据使用、生物样本保存与销毁 | Pass / Fail | - | GCP + 生物安全法 |
| 14 | 受试者可获取研究结果总结 | Pass / Fail | - | 赫尔辛基宣言 |
| 15 | 明确注明:这是研究,不是常规治疗 | Pass / Fail | - | GCP + IRB要求 |
| 16 | 禁止出现强迫、诱导、免责不公平条款 | Pass / Fail | - | GCP + ICH-GCP 4.8.11 |
**判定依据:**
- 16条全部Pass → **同意** ✓
- 任意一条Fail → **修改后同意** ✗
- 8条及以上Fail → **不同意** ✗✗
---
## 总体结论
**[同意 / 修改后同意 / 不同意]**
---
## 缺失强制条款列表
(仅当存在Fail项时填写)
| # | 条款编号 | 缺失内容 | 相关法规 | 严重程度 |
|---|---------|---------|---------|---------|
| 1 | xxx | xxx | xxx | 高/中/低 |
---
## 不合规表述
(仅当发现不合规表述时填写)
| # | 原文表述 | 条款编号 | 问题描述 | 修改建议 |
|---|---------|---------|---------|---------|
| 1 | xxx | xxx | xxx | xxx |
---
## 伦理风险点
### 高风险
- xxx
### 中风险
- xxx
### 低风险
- xxx
---
## 必须修改内容(优先处理)
1. xxx
2. xxx
---
## 建议优化内容(提升质量)
1. xxx
2. xxx
---
## 参考法规条款
- 《药物临床试验质量管理规范》(2020版)第XX条
- ICH-GCP E6(R2) 第4.8条
- 21 CFR Part 50 第50.25条
- 世界医学大会赫尔辛基宣言(2013版)
语言规范性要求
- ✓ 通俗易懂,避免专业术语(必要时提供通俗解释)
- ✓ 量化表述(如"约XX%""可能出现XX次")
- ✗ 禁止"一定""保证""绝对有效""治愈"等承诺
- ✗ 禁止淡化风险(如"无任何副作用")
- ✗ 禁止夸大受益(如"肯定有效""最佳治疗")
- ✗ 禁止使用免责不公平条款
特殊人群保护(额外审查)
以下人群参与研究时,需额外审查保护措施:
- 儿童(未成年人):须监护人同意 + 儿童理解能力适配说明
- 孕妇/哺乳期妇女:对胎儿/乳儿的额外风险说明
- 精神障碍/认知受损者:法定代理人同意 + 受试者本人意愿确认
- 弱势群体(学生、员工、军人等):特别注意无诱导、无强迫
参考文件
技能包中包含以下参考文件,审查时请结合使用:
REFERENCES/FDA_Informed_Consent_Guidance_2023.md— FDA 2023年知情同意指南(权威参考)REFERENCES/ICF_Elements_Checklist.md— ICF条款核查清单(快速对照)EXAMPLES/Sample_ICF_Report.md— 审查报告示例(格式参考)
Usage Guidance
This skill appears coherent and focused on reviewing informed-consent forms. Before installing/using it: 1) Confirm you trust the author/source (homepage unknown). 2) Note the scripts call an external tool named 'markitdown' — ensure you have the expected tool installed from a trusted source to avoid executing an unexpected binary. 3) The scripts are interactive (read -p) and expect local file access; do not run them on systems with untrusted or highly sensitive patient data unless your environment complies with your institutional privacy rules. 4) The manifest/README contains small filename mismatches — verify paths before running. If you need autonomous agent invocation to process sensitive ICF PDFs, consider whether automated access to patient data is allowed by your policies; this skill itself does not exfiltrate data, but any agent you grant file-access to could transmit content if configured elsewhere.
Capability Analysis
Type: OpenClaw Skill
Name: icf-review
Version: 1.0.0
The icf-review skill bundle is a legitimate tool designed for regulatory compliance auditing of clinical trial Informed Consent Forms (ICF). The scripts (icf_review_workflow.sh and pdf_to_icf_md.sh) perform standard file conversion and text processing using the 'markitdown' utility, and the SKILL.md provides structured, domain-specific instructions for an AI agent to act as a GCP/IRB auditor without any evidence of prompt injection, data exfiltration, or malicious execution.
Capability Tags
Capability Assessment
Purpose & Capability
Name, description, SKILL.md, example report, checklist and reference documents are all aligned with performing informed-consent (ICF) compliance reviews. The included scripts are small helper utilities for converting PDFs to markdown and guiding a manual review — these are reasonable for this skill.
Instruction Scope
SKILL.md and scripts stay within the stated purpose (convert ICF files, run a manual/structured review, produce a report). Minor inconsistencies: the scripts and workflow text reference filenames that do not exactly match the manifest (e.g., script mentions REFERENCES/FDA_ICF_Guidance.pdf and EXAMPLES/ICF_Review_Report_Example.md, while the manifest contains REFERENCES/FDA_Informed_Consent_Guidance_2023.md and EXAMPLES/Sample_ICF_Report.md). Also the conversion script invokes an external CLI tool 'markitdown' which is not declared as a required binary or installed by the package; this is an operational dependency (not a secret-exfiltration risk) but may cause failures or unexpected behavior if a different 'markitdown' binary exists on the host.
Install Mechanism
No install spec (instruction-only with small local scripts) — lowest-risk model. Nothing is downloaded from external URLs, no archives are extracted, and no package registries are pulled during install.
Credentials
The skill declares no required environment variables, no credentials, and no config paths. The runtime instructions and scripts do not attempt to read secrets or unrelated system configuration.
Persistence & Privilege
always:false and no code tries to modify other skills or global agent settings. The skill is user-invocable only and does not request elevated or persistent privileges.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install icf-review - After installation, invoke the skill by name or use
/icf-review - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
临床试验知情同意书(ICF)全面审查 — 基于GCP/ICH-GCP/FDA/中国法规,对研究项目的知情同意书进行合规性、完整性、伦理合理性审查,并输出结构化审查报告。触发词:审查知情同意书、ICF审查、IRB审查、知情同意审查。
Metadata
Frequently Asked Questions
What is 临床研究知情同意书审查?
临床试验知情同意书(ICF)全面审查 — 基于GCP/ICH-GCP/FDA/中国法规,对知情同意书进行合规性、完整性、伦理合理性审查,输出结构化审查报告。触发词:审查知情同意书、ICF审查、IRB审查、知情同意审查。 It is an AI Agent Skill for Claude Code / OpenClaw, with 69 downloads so far.
How do I install 临床研究知情同意书审查?
Run "/install icf-review" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is 临床研究知情同意书审查 free?
Yes, 临床研究知情同意书审查 is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does 临床研究知情同意书审查 support?
临床研究知情同意书审查 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created 临床研究知情同意书审查?
It is built and maintained by kenlcj (@kenlcj); the current version is v1.0.0.
More Skills