AgentSkillsCN

Research Project Scaffold

研究项目框架

SKILL.md

Research Project Scaffold Skill

Scaffolding new academic research projects and refactoring existing ones for Alex-assisted workflows.

Purpose

Provide standardized structures and refactoring procedures for academic research projects that maximize Alex's ability to assist throughout the research lifecycle.


Scaffolding New Projects

Recommended Folder Structure

code
project-root/
├── .github/
│   ├── copilot-instructions.md    # Research-specific Alex context
│   └── prompts/
│       └── literature-synthesis.prompt.md
├── docs/
│   ├── RESEARCH-PLAN.md           # Objectives, timeline, milestones
│   ├── METHODOLOGY.md             # Research design decisions
│   └── DECISION-LOG.md            # Key decisions with rationale
├── data/
│   ├── raw/                       # Untouched source data
│   ├── processed/                 # Cleaned/transformed data
│   └── DATA-DICTIONARY.md         # Variable definitions
├── analysis/
│   ├── scripts/                   # Analysis code
│   ├── outputs/                   # Generated figures, tables
│   └── notebooks/                 # Exploratory analysis
├── writing/
│   ├── drafts/                    # Work in progress
│   ├── figures/                   # Publication-ready figures
│   └── submissions/               # Submitted versions
├── references/
│   ├── LITERATURE-MATRIX.md       # Systematic literature tracking
│   ├── pdfs/                      # Source papers (if permitted)
│   └── notes/                     # Reading notes
└── README.md                      # Project overview

Essential Files

RESEARCH-PLAN.md Template

markdown
# Research Plan: [Title]

## Research Questions
1. Primary: [Main question]
2. Secondary: [Supporting questions]

## Methodology
- Type: [Qualitative/Quantitative/Mixed/Case Study]
- Design: [Experimental/Survey/Longitudinal/etc.]
- Participants: [Sample description]

## Timeline
| Phase | Duration | Deliverable |
|-------|----------|-------------|
| Literature Review | Weeks 1-4 | LITERATURE-MATRIX.md complete |
| Data Collection | Weeks 5-12 | Raw data in data/raw/ |
| Analysis | Weeks 13-16 | Results in analysis/outputs/ |
| Writing | Weeks 17-20 | Draft in writing/drafts/ |
| Revision | Weeks 21-24 | Submission-ready manuscript |

## Success Criteria
- [ ] [Specific measurable outcome]
- [ ] [Publication target]
- [ ] [Other goals]

## Risks & Mitigations
| Risk | Likelihood | Mitigation |
|------|------------|------------|
| [Risk 1] | Medium | [Plan] |

LITERATURE-MATRIX.md Template

markdown
# Literature Matrix

## Search Strategy
- Databases: [List sources]
- Keywords: [Search terms]
- Inclusion criteria: [What qualifies]
- Exclusion criteria: [What's out]

## Matrix

| Citation | Year | Methodology | Key Findings | Gaps | Relevance | Notes |
|----------|------|-------------|--------------|------|-----------|-------|
| Author1 et al. | 2024 | Controlled study | Finding X | Gap Y | ⭐⭐⭐ | Foundational |
| Author2 | 2023 | Survey (n=500) | Finding Z | — | ⭐⭐ | Supports H2 |

## Synthesis Notes
- Theme 1: [Pattern across sources]
- Theme 2: [Another pattern]
- Gap: [What's missing that your research addresses]

METHODOLOGY.md Template

markdown
# Methodology

## Research Design
**Approach**: [Paradigm - positivist, interpretivist, pragmatic]
**Strategy**: [Case study, experiment, survey, ethnography, etc.]

## Data Collection
**Sources**: [What data, from where]
**Instruments**: [Surveys, interviews, logs, artifacts]
**Procedure**: [Step-by-step process]

## Analysis Method
**Technique**: [Thematic analysis, statistical tests, content analysis]
**Tools**: [Software, frameworks]
**Validation**: [Member checking, triangulation, statistical power]

## Ethical Considerations
- IRB Status: [Approved/Exempt/N/A]
- Consent: [Process]
- Data protection: [How handled]

## Limitations
1. [Known limitation and why acceptable]
2. [Another limitation]

copilot-instructions.md Template (Research Projects)

markdown
# [Project Title] - Research Context

## Project Overview
[2-3 sentence summary of research focus and goals]

## Current Phase
- [x] Literature Review
- [ ] Data Collection  
- [ ] Analysis
- [ ] Writing

## Key Files
- Research plan: docs/RESEARCH-PLAN.md
- Literature: references/LITERATURE-MATRIX.md
- Methodology: docs/METHODOLOGY.md

## Alex Guidance
- When reviewing literature: Add entries to LITERATURE-MATRIX.md
- When analyzing data: Document steps in DECISION-LOG.md
- When writing: Check references against LITERATURE-MATRIX.md
- Citation style: [APA 7 / Chicago / IEEE / etc.]

## Domain Context
[Key concepts, terminology, or background Alex needs]

## Quality Standards
- All claims require citations or data references
- Methodology decisions logged with rationale
- Raw data never modified (work from processed/)

Refactoring Existing Projects

Assessment Checklist

Run this audit on any existing research project:

markdown
## Research Project Audit

### Structure Assessment
- [ ] Has dedicated .github/ folder with copilot-instructions.md
- [ ] Data separated: raw vs processed
- [ ] Clear separation: analysis code vs writing
- [ ] References organized and tracked

### Documentation Assessment  
- [ ] Research questions documented
- [ ] Methodology decisions recorded
- [ ] Literature systematically tracked (not scattered notes)
- [ ] Decision rationale captured

### Alex-Readiness Assessment
- [ ] Context file exists (copilot-instructions.md or equivalent)
- [ ] Current phase clearly marked
- [ ] Key terminology defined
- [ ] Citation style specified

### Data Management Assessment
- [ ] Raw data preserved unchanged
- [ ] Data dictionary exists
- [ ] Analysis reproducible
- [ ] Sensitive data protected

Refactoring Procedure

Phase 1: Inventory (Non-Destructive)

  1. Create inventory file:

    markdown
    # Refactoring Inventory - [Project Name]
    
    ## Current Structure
    [Tree output of existing folders]
    
    ## Identified Files by Type
    ### Writing/Drafts
    - file1.docx - [description]
    
    ### Data
    - dataset.csv - [raw/processed?]
    
    ### Analysis
    - script.py - [purpose]
    
    ### References
    - [scattered papers, notes]
    
    ### Unknown/Misc
    - [files needing classification]
    
  2. Identify gaps:

    • Missing: RESEARCH-PLAN? METHODOLOGY? LITERATURE-MATRIX?
    • Undocumented: Key decisions made but not recorded?
    • Scattered: References in multiple locations?

Phase 2: Scaffold Creation

  1. Create target structure (empty folders):

    code
    mkdir .github docs data data/raw data/processed analysis writing references
    
  2. Create missing essential files:

    • RESEARCH-PLAN.md (even if retrospective)
    • METHODOLOGY.md
    • LITERATURE-MATRIX.md
    • copilot-instructions.md
  3. Populate retrospectively:

    • Extract research questions from drafts
    • Document methodology from what was done
    • Build literature matrix from existing citations

Phase 3: File Migration

Principle: Move, don't copy (avoid duplicates). Use git.

bash
# Example migration commands
git mv old-data.csv data/raw/
git mv cleaned-data.csv data/processed/
git mv draft-v3.docx writing/drafts/
git mv analysis.py analysis/scripts/

Migration mapping template:

Original LocationNew LocationAction
data.csvdata/raw/original-data.csvRename + move
data_cleaned.csvdata/processed/data-cleaned.csvMove
paper.docxwriting/drafts/paper-v1.docxVersion + move
notes.txtreferences/notes/reading-notes.mdConvert + move

Phase 4: Retroactive Documentation

Fill in documentation from what exists:

  1. RESEARCH-PLAN.md: Extract from proposal, drafts, or reconstruct
  2. DECISION-LOG.md: Interview yourself about choices made
  3. LITERATURE-MATRIX.md: Build from reference list in drafts
  4. DATA-DICTIONARY.md: Document variables from analysis code

Retroactive documentation prompts:

  • "What was I trying to answer?"
  • "Why did I choose this method over alternatives?"
  • "What did Author X contribute to my thinking?"
  • "What does this variable actually represent?"

Phase 5: Link Updates

  1. Update any internal links in documents
  2. Update file paths in analysis scripts
  3. Update .gitignore for new structure
  4. Update any build/compile scripts

Phase 6: Validation

  • All files accounted for (nothing lost)
  • Analysis scripts run from new locations
  • Links in documents work
  • Git history preserved (used mv, not delete+create)
  • README.md updated with new structure

Best Practices

By Research Phase

PhaseAlex IntegrationKey Actions
PlanningResearch design consultationCreate RESEARCH-PLAN.md, set up structure
LiteratureSummarization, gap analysisBuild LITERATURE-MATRIX.md systematically
Data CollectionInstrument review, protocol checkingDocument in METHODOLOGY.md
AnalysisCode review, interpretation checkingLog decisions, version scripts
WritingDrafting, citation checking, editingUse references/ for fact-checking
RevisionPeer review response, formattingTrack versions in writing/submissions/

Common Refactoring Patterns

SmellRefactoring
All files in rootCreate folder hierarchy
Data mixed with analysisSeparate data/raw, data/processed, analysis/
Multiple draft versions scatteredConsolidate in writing/drafts with clear naming
References in random locationsCentralize in references/
No context for AI assistanceCreate copilot-instructions.md
Decisions made but not recordedCreate retrospective DECISION-LOG.md
Literature notes disorganizedBuild structured LITERATURE-MATRIX.md

File Naming Conventions

code
# Data files
data/raw/survey-responses-2024-01-15.csv     # Source + date
data/processed/survey-clean-v2.csv           # Description + version

# Writing
writing/drafts/paper-draft-01.md             # Numbered versions
writing/submissions/journal-name-2024-03.pdf # Venue + date

# Analysis
analysis/scripts/01-data-cleaning.py         # Numbered sequence
analysis/scripts/02-descriptive-stats.py
analysis/outputs/figure-1-demographics.png   # Matches paper reference

Version Control Practices

  • Commit at logical checkpoints (completed section, working analysis)
  • Use meaningful commit messages: "Add literature synthesis for RQ1"
  • Tag important versions: git tag submission-v1
  • Never commit sensitive data (use .gitignore)

Integration with Other Skills

  • academic-research: Use for methodology guidance, committee navigation
  • practitioner-research: Use for publication workflows (Ship→Document→Promote)
  • project-scaffolding: Base patterns for folder structure
  • bootstrap-learning: For learning new research methods
  • writing-publication: For manuscript preparation

Clinical Research Extension

Patterns from Lithium Alzheimer's prevention research project (Feb 2026)

Clinical Research Folder Structure

code
project-root/
├── .github/
│   └── copilot-instructions.md
├── clinical/
│   ├── IRB-PROTOCOL.md           # Full IRB submission document
│   ├── IRB-SELECTION-GUIDE.md    # IRB comparison if no internal IRB
│   ├── INFORMED-CONSENT.md       # English consent form
│   ├── INFORMED-CONSENT-ES.md    # Spanish translation
│   ├── INFORMED-CONSENT-PT.md    # Portuguese translation
│   ├── CASE-REPORT-FORMS.md      # Data collection instruments
│   ├── SCREENING-CHECKLIST.md    # Eligibility criteria forms
│   ├── FUNDING-STRATEGY.md       # Grant/funding applications
│   └── MASTER-CHECKLIST.md       # Project-wide task tracker
├── literature/
│   ├── LITERATURE-MATRIX.md      # Systematic evidence tracking
│   ├── PRIMARY-SOURCES.md        # Key papers with PMIDs
│   ├── CLINICAL-TRIALS.md        # Existing trial summaries
│   └── EPIDEMIOLOGICAL-STUDIES.md # Population-level evidence
├── research/
│   └── RESEARCH-PROTOCOL.md      # Scientific methodology
├── analysis/
│   ├── EVIDENCE-SYNTHESIS.md     # Meta-analysis results
│   └── HYPOTHESIS-EVALUATION.md  # Systematic claim validation
├── docs/
│   ├── EXECUTIVE-SUMMARY.md      # Leadership/stakeholder brief
│   ├── CLINICAL-IMPLICATIONS.md  # Provider guidance
│   ├── STUDY-GOALS.md            # Research objectives
│   └── FUTURE-RESEARCH.md        # Next steps, Phase 2+
└── README.md                     # 3-minute pitch + overview

IRB Protocol Template Sections

  1. Protocol Information — Title, version, sponsor, IND status
  2. Principal Investigators — Names, credentials, roles, contacts
  3. Study Synopsis — Background, rationale, objectives, design
  4. Study Population — Inclusion/exclusion criteria with rationale
  5. Study Intervention — What is being tested, dosing, duration
  6. Study Procedures — Visit schedule, assessments, data collection
  7. Safety Monitoring — Adverse event definitions, reporting, stopping rules
  8. Data Management — Collection, storage, HIPAA compliance
  9. Statistical Analysis — Power calculation, analysis plan
  10. Ethical Considerations — Risk/benefit, consent process, vulnerable populations

Inclusion/Exclusion Criteria Pattern

markdown
### Inclusion Criteria
1. Age ≥ [minimum] years
2. [Clinical condition or symptom]
3. [Baseline test values] (e.g., eGFR > 60)
4. Ability to provide informed consent
5. Willing to comply with study procedures

### Exclusion Criteria
1. **Organ impairment**: [specific threshold]
2. **Current medication use**: [contraindicated drugs]
3. **Comorbid conditions**: [list with rationale]
4. **Concurrent medications** due to interaction risk:
   - [Drug class 1] (rationale)
   - [Drug class 2] (rationale)
5. **Pregnancy or planning pregnancy**
6. **Life expectancy < [study duration]**

External IRB Selection (Non-Academic Settings)

IRBCost RangeTimelineBest For
Sterling IRB$1,500-3,0002-3 weeksSupplement studies
Solutions IRB$1,000-2,5002-3 weeksBudget-conscious
Advarra$2,000-4,0002-4 weeksAcademic spinoffs
WCG IRB$2,500-5,0002-4 weeksLarge commercial

Budget: Add $2,000-3,000 for IRB fees + 4-6 weeks to timeline.

Evidence Confidence Levels

LevelMeaningBasis
HIGHStrong consensusMultiple RCTs, mechanisms confirmed
MODERATEPromisingAnimal studies, small human trials
LOWEmergingCase reports, theoretical basis only
INSUFFICIENTUnknownNo quality evidence available

Stakeholder Materials

  • README.md — 3-minute pitch for decision-makers
  • EXECUTIVE-SUMMARY.md — 1-page BLUF for busy executives
  • CLINICAL-IMPLICATIONS.md — Provider-focused guidance

Clinical Research Localization

For multi-language studies, maintain parallel informed consent documents. Use standard suffixes:

LanguageSuffixExample
English(none)INFORMED-CONSENT.md
Spanish-ESINFORMED-CONSENT-ES.md
Portuguese-PTINFORMED-CONSENT-PT.md
Chinese-ZHINFORMED-CONSENT-ZH.md

Quick Commands

code
# New project setup
mkdir -p .github/prompts docs data/{raw,processed} analysis/{scripts,outputs} writing/{drafts,figures} references/notes
touch docs/RESEARCH-PLAN.md docs/METHODOLOGY.md docs/DECISION-LOG.md references/LITERATURE-MATRIX.md .github/copilot-instructions.md README.md

# Audit existing project
find . -type f -name "*.md" -o -name "*.csv" -o -name "*.py" | head -50

# Migration with git tracking
git mv source destination

Troubleshooting

IssueSolution
Project too messy to auditStart with just essential files, refactor incrementally
Lost track of versionsUse git log, create retrospective changelog
Data too sensitive for repoUse .gitignore, document in DATA-DICTIONARY.md what exists
Analysis scripts break after moveUpdate relative paths, consider using project root as working directory
Can't remember decisionsCheck email, drafts, commit messages for clues

This skill makes research projects Alex-ready from day one, or gets them there through systematic refactoring.