framework-guided-formulation

$npx mdskill add yogsoth-ai/de-anthropocentric-research-engine/framework-guided-formulation

Uses research frameworks to structure and refine research questions systematically

  • Solves the problem of formulating incomplete or vague research questions
  • Depends on context-management and subagent-spawning for task coordination
  • Chooses PICO, SPIDER, SPICE, or ECLIPSE based on research type and criteria
  • Delivers structured RQs by filling framework components with specific details

SKILL.md

.github/skills/framework-guided-formulationView on GitHub ↗
---
name: framework-guided-formulation
description: "Strategy: 选择 RQ 框架(PICO/SPIDER/SPICE/ECLIPSE)并系统应用"
version: 1.0.0
category: hypothesis-formation
type: strategy
campaign: research-question
tactics:
  - framework-selection-and-application
  - question-refinement-loop
sops:
  - framework-matching
  - pico-application
  - spider-application
  - spice-application
  - eclipse-application
  - finer-criteria-check
dependencies:
  skills:
    - context-management
    - subagent-spawning
---

# Framework-Guided Formulation

选择最适合的 RQ 框架并系统应用 — 当研究类型明确时,用标准框架结构化问题。

## 适用场景

- 研究类型明确(量化/定性/混合/评估)
- 有对应的标准框架可用
- 需要确保问题的各组件完整

## 思维框架

核心逻辑: 不同研究类型有不同的"好问题"标准。框架是前人总结的"好问题应该包含哪些组件"的清单。选对框架 → 填充每个组件 → 自然得到结构化的问题。

### 框架选择指南

| 研究类型 | 推荐框架 | 核心组件 |
|----------|---------|---------|
| 量化/干预研究 | PICO/PICOTS | Population, Intervention, Comparison, Outcome (+Time, Setting) |
| 定性研究 | SPIDER | Sample, Phenomenon of Interest, Design, Evaluation, Research type |
| 评估研究 | SPICE | Setting, Perspective, Intervention, Comparison, Evaluation |
| 混合方法 | ECLIPSE | Expectation, Client group, Location, Impact, Professionals, Service |

### 框架应用原则

- 每个组件必须有明确填充(不留空)
- 填充内容必须具体(不是"相关人群"而是"18-65岁的2型糖尿病患者")
- 如果某组件不适用,显式说明为什么不适用

## Budget Gate

| Tier | 框架候选 | 框架应用 | FINER 检验 | 产出 |
|------|---------|---------|-----------|------|
| S | ≥2 框架比较 | 选定框架全组件填充 | 5 项全通过 | ≥1 RQ |
| M | ≥3 框架比较 | 选定框架全组件 + 替代框架对比 | 5 项 + success criteria | ≥2 RQ |
| L | ≥4 框架比较 | 多框架并行应用 + 最优选择 | 5 项 + criteria + 敏感性 | ≥3 RQ |

## 默认参考流

1. 确定研究类型(量化/定性/混合/评估)
2. 匹配候选框架(framework-matching SOP)
3. 应用选定框架(pico/spider/spice/eclipse-application SOP)
4. FINER 检验(finer-criteria-check SOP)
5. 未通过项修正 → 重新检验
6. 产出最终 RQ

## context-checkpoint

Strategy 完成后必须调用 context-checkpoint,记录:
- 选定的框架及理由
- 框架各组件的填充内容
- FINER 检验结果
- 最终 RQ 表述

More from yogsoth-ai/de-anthropocentric-research-engine