Skip to content
Back to skills

Systematic Review

ASecurity

Run a PRISMA 2020-compliant systematic literature review with optional meta-analysis — protocol registration, systematic search with documented strategy, screening with inclusion/exclusion criteria, risk-of-bias assessment (RoB 2 / ROBINS-I), effect-size synthesis with heterogeneity and GRADE, plus PRISMA-trAIce AI-transparency checklist and RAISE responsible-AI reporting. Use when the user asks for a systematic review, meta-analysis, PRISMA-compliant evidence synthesis, or a survey that must...

  • 9 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 7, 2026
researchapi

Works with

  • api
  • mcp

Security analysis

A100/100

Pro scans all 5 files and shows the line behind each finding

Scanned October 7, 2026

npx -y skills add Yichen-ZJU/claude-for-research --skill systematic-review --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Systematic Review?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Systematic Review
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/yichen-zju-systematic-review/badge)](https://www.skillsdirectory.com/skills/yichen-zju-systematic-review)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: systematic-review
description: Run a PRISMA 2020-compliant systematic literature review with optional meta-analysis — protocol registration, systematic search with documented strategy, screening with inclusion/exclusion criteria, risk-of-bias assessment (RoB 2 / ROBINS-I), effect-size synthesis with heterogeneity and GRADE, plus PRISMA-trAIce AI-transparency checklist and RAISE responsible-AI reporting. Use when the user asks for a systematic review, meta-analysis, PRISMA-compliant evidence synthesis, or a survey that must meet formal reporting standards (as opposed to a narrative literature review).
argument-hint: <research-question>
---

# Systematic Review(PRISMA 系统性综述)

做符合 PRISMA 2020 规范的系统性综述/荟萃分析。**和 `literature-review`(叙述性综述)的区别**:本 skill 产出可注册、可复现、有合规报告的正式证据合成——搜索策略、纳排标准、偏倚评估全部留痕。

参考协议(本 skill references/ 下,出处见文末):
- `prisma_trAIce_protocol.md` — AI 透明报告清单(JMIR 2025),Mandatory 项不过即阻断
- `raise_framework.md` — 证据合成中负责任使用 AI(NIHR/UCL)
- `systematic_review_protocol.md` — 五阶段流程与检查点
- `systematic_review_toolkit.md` — Cochrane 手册、RoB 2、ROBINS-I、I²、GRADE、协议注册

## 五阶段流程

```
Phase 1  协议制定与注册
         RQ(PICO/SPIDER 框架)→ 检索策略(数据库、关键词、布尔式)
         → 纳入/排除标准 → 写 protocol(templates 见 toolkit)
         → 可注册 OSF/PROSPERO。先与用户确认 protocol 再动手
Phase 2  系统检索
         按 protocol 执行(arxiv MCP + web_search + Semantic Scholar API)
         记录每个数据库的检索式、日期、命中数 → PRISMA 流程图数据
Phase 3  筛选与选择
         标题/摘要初筛 → 全文复筛,每一步记录排除理由和数量
         更新 PRISMA flow: identified → screened → assessed → included
Phase 4  数据提取与偏倚评估
         提取表(作者/年/方法/样本/效应量/结论)
         RoB 2(随机试验)或 ROBINS-I(非随机)逐篇评估,红绿灯汇总
Phase 5  综合与报告
         可定量 → meta 分析(效应量、异质性 I²、敏感性分析、GRADE 证据分级)
         不可定量 → 叙述性综合(也要 GRADE 自评)
         PRISMA 2020 报告(27 项清单)+ trAIce AI 使用清单 + RAISE 角色矩阵
```

## 检查点(阻断性)

1. Protocol 未经用户确认 → 不得开始检索
2. PRISMA-trAIce **Mandatory** 项未满足 → 阻断交付
3. 检索策略、纳排标准、分析方法的文档缺失 → 视为未完成(可复现性是硬要求)
4. 灰区文献(无法确认存在)= FAIL,不入报告(同全局 provenance 约定)

## 产出与落盘

slug 约定:`outputs/<slug>-sr-protocol.md`、`outputs/<slug>-sr-flow.md`、`outputs/<slug>-sr-extraction.md`、`outputs/<slug>-sr.md`(最终报告)+ `<slug>.provenance.md`。meta 分析的原始数据(效应量表、I² 计算脚本)必须保留。

## 合规披露

最终报告必须包含:AI 使用声明(哪些环节用了 AI、人工如何核验)——用 `ai-use-disclosure` skill 生成。检索日期、数据库清单、各库命中数。

## 出处

协议文件改编自 ARS (academic-research-skills, Cheng-I Wu, CC-BY-NC 4.0)、PRISMA-trAIce (Holst et al., JMIR AI 2025, doi:10.2196/80247)、RAISE (Thomas et al., NIHR ESG BPWG, 2025)。

Files in this skill

  • SKILL.md3.5 KB
  • references/prisma_trAIce_protocol.md10.2 KB
  • references/raise_framework.md7.2 KB
  • references/systematic_review_protocol.md3.9 KB
  • references/systematic_review_toolkit.md18.7 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.