project-sharing（项目分享）

Project Sharing and Output Preparation

Expert guidance for preparing project outputs for sharing with collaborators, reviewers, or repositories. Creates organized packages at different sharing levels while preserving your working directory.

Supporting files in this directory:

• notebook-streamlining.md - Streamlining notebooks for sharing and the abridge option
• quality-assurance.md - QA procedures, best practices, checklists, and dependency management
• common-scenarios.md - Sharing scenarios (collaborators, manuscripts, archival, repositories) and example scripts
• cleanup-and-deprecation.md - Correcting cleanup mistakes and deprecating redundant notebooks

When to Use This Skill

• Sharing analysis results with collaborators
• Preparing supplementary materials for publications
• Creating reproducible research packages
• Archiving completed projects
• Handoff to other researchers
• Submitting to data repositories

Core Principles

1. Work on copies - Never modify the working directory
2. Choose appropriate level - Match sharing depth to audience needs
3. Document everything - Include clear guides and metadata
4. Clean before sharing - Remove debug code, clear outputs, anonymize if needed
5. Make it reproducible - Include dependencies and instructions
6. CRITICAL: After creating sharing folder, all future work happens in the main project directory, NOT in the sharing folder - Sharing folders are read-only snapshots

Three Sharing Levels

Level 1: Summary Only

Purpose: Quick sharing for presentations, reports, or high-level review

What to include:

• PDF export of final notebook(s)
• Final data/results (CSV, Excel, figures) - optional
• Brief README

Use when:

• Sharing results with non-technical stakeholders
• Presentations or talks
• Quick review without reproduction needs
• Space/time constraints

Structure:

shared-summary/
├── README.md                          # Brief overview
├── analysis-YYYY-MM-DD.pdf           # Notebook as PDF
└── results/
    ├── figures/
    │   ├── fig1-main-result.png
    │   └── fig2-comparison.png
    └── tables/
        └── summary-statistics.csv

Level 2: Reproducible

Purpose: Enable others to reproduce your analysis from processed data

What to include:

• Analysis notebooks (.ipynb) - cleaned
• Scripts for figure generation
• Processed/analysis-ready data
• Requirements file (requirements.txt or environment.yml)
• Detailed README with instructions

Use when:

• Sharing with collaborating researchers
• Peer review / manuscript supplementary materials
• Teaching or tutorials
• Standard collaboration needs

Structure:

For standard project structures, see the folder-organization skill. Reproducible packages should include:

• Processed data (in data/processed/)
• Cleaned notebooks (in notebooks/) with outputs cleared
• Scripts (in scripts/)
• Environment specification (environment.yml or requirements.txt)
• Documentation (README.md, MANIFEST.md)

shared-reproducible/
├── README.md                          # Setup and reproduction instructions
├── MANIFEST.md                        # File descriptions
├── environment.yml                    # Conda environment OR requirements.txt
├── notebooks/                         # Cleaned notebooks
├── scripts/                           # Standalone scripts
└── data/
    └── processed/                     # Analysis-ready data

Level 3: Full Traceability

Purpose: Complete transparency from raw data through all processing steps

What to include:

• Starting/raw data
• All processing scripts and notebooks
• All intermediate files
• Final results
• Complete documentation
• Full dependency specification

Use when:

• Archiving for future reference
• Regulatory compliance
• High-stakes reproducibility (clinical, policy)
• Data repository submission (Zenodo, Dryad, etc.)
• Complete project handoff

Structure:

For standard project structures, see the folder-organization skill. Full traceability packages should include complete data hierarchy:

shared-complete/
├── README.md                          # Complete project guide
├── MANIFEST.md                        # Comprehensive file listing
├── environment.yml
├── data/
│   ├── raw/                          # Original, unmodified data
│   ├── intermediate/                 # Processing steps
│   └── processed/                    # Final analysis-ready
├── scripts/                           # All processing scripts
├── notebooks/                         # All notebooks (exploratory + final)
├── results/                           # All outputs
│   ├── figures/
│   ├── tables/
│   └── supplementary/
└── documentation/                     # Complete documentation
    ├── methods.md
    ├── changelog.md
    └── data-dictionary.md

Preparation Workflow

Step 1: Ask User for Sharing Level

Questions to determine level:

Which sharing level do you need?

1. Summary Only - PDF + final results (quick sharing)
2. Reproducible - Notebooks + scripts + data (standard sharing)
3. Full Traceability - Everything from raw data (archival/compliance)

Additional questions:
- Who is the audience? (colleagues, reviewers, public)
- Are there size constraints?
- Any sensitive data to handle?
- Timeline for sharing?

Step 2: Identify Files to Include

Level 1 - Summary:

• Main analysis notebook(s)
• Key figures (publication-quality)
• Summary tables/statistics

Level 2 - Reproducible:

• All analysis notebooks (not exploratory)
• Figure generation scripts
• Processed/cleaned data
• Environment specification
• Any utility functions/modules

Level 3 - Full:

• Raw data (or links if too large)
• All processing scripts
• All notebooks (including exploratory)
• All intermediate files
• Complete documentation

Step 3: Create Sharing Directory

# Create dated directory
SHARE_DIR="shared-$(date +%Y%m%d)-[level]"
mkdir -p "$SHARE_DIR"

Step 4: Copy and Clean Files

For notebooks (.ipynb):

import nbformat
from nbconvert.preprocessors import ClearOutputPreprocessor

def clean_notebook(input_path, output_path):
    """Clean notebook: clear outputs, remove debug cells."""
    with open(input_path, 'r') as f:
        nb = nbformat.read(f, as_version=4)

    clear_output = ClearOutputPreprocessor()
    nb, _ = clear_output.preprocess(nb, {})

    nb.cells = [cell for cell in nb.cells
                if 'debug' not in cell.metadata.get('tags', [])
                and 'remove' not in cell.metadata.get('tags', [])]

    with open(output_path, 'w') as f:
        nbformat.write(nb, f)

For data files: Copy as-is for small files; compress large files; check for sensitive information.

For scripts: Remove debugging code; add docstrings if missing; ensure paths are relative.

For notebook streamlining and the abridge option, see notebook-streamlining.md.

Step 4.5: Verify and Fix File Paths

Problem: Notebooks and scripts with broken file paths will fail when shared.

For complete path verification procedures, automated checking scripts, and correction patterns, see the folder-organization skill.

Quick check commands:

# Check for absolute paths in notebooks
grep -l "/Users/" *.ipynb
grep -l "C:\\\\" *.ipynb

Step 5: Generate Documentation

README.md Template

# Project: [Project Name]

**Date:** YYYY-MM-DD
**Author:** [Your Name]
**Sharing Level:** [Summary/Reproducible/Full]

## Overview
Brief description of the project and analysis.

## Contents
See MANIFEST.md for detailed file descriptions.

## Requirements
[For Reproducible/Full levels]
- Python 3.X
- See environment.yml for dependencies

## Setup
\`\`\`bash
conda env create -f environment.yml
conda activate project-name
\`\`\`

## Reproduction Steps
[For Reproducible/Full levels]
1. [Description of first step]

## Data Sources
[For Full level]
- Dataset A: [Source, download date, version]

## Contact
[Your email or preferred contact]

## License
[If applicable - e.g., CC BY 4.0, MIT]

MANIFEST.md Template

# File Manifest
Generated: YYYY-MM-DD

## File Descriptions

### Notebooks
- \`notebooks/01-data-processing.ipynb\` - Initial data loading and cleaning
- \`notebooks/02-analysis.ipynb\` - Main statistical analysis

### Data
- \`data/processed/cleaned_data.csv\` - Quality-controlled dataset (N=XXX samples)

### Scripts
- \`scripts/generate_figures.py\` - Automated figure generation

### Results
- \`results/figures/fig1-main.png\` - Main result showing [description]

Step 6: Handle Sensitive Data

Check for: PII, access credentials, proprietary data, institutional restrictions, patient/subject identifiers.

Strategies:

1. Anonymize - Remove or hash identifiers
2. Exclude - Don't include sensitive files
3. Aggregate - Share summary statistics only
4. Document restrictions - Note what's excluded and why

Step 7: Package and Compress

# For smaller packages (<100MB)
zip -r shared-YYYYMMDD.zip shared-YYYYMMDD/

# For larger packages
tar -czf shared-YYYYMMDD.tar.gz shared-YYYYMMDD/

Step 8: Return to Working Directory

IMPORTANT: After creating the sharing package, always work in the main project directory.

The sharing folder is a snapshot for distribution only. Any future development, analysis, or modifications should happen in your original working directory.

cd /path/to/main/project  # Return to working directory
pwd                        # Verify location
# Continue work here, NOT in shared-YYYYMMDD/

Integration with Other Skills

Works well with:

• folder-organization - Ensures source project is well-organized before sharing
• jupyter-notebook-analysis - Creates notebooks that are share-ready
• managing-environments - Documents dependencies properly

Before using this skill:

1. Organize working directory (folder-organization)
2. Finalize analysis (jupyter-notebook-analysis)
3. Document environment (managing-environments)

After using this skill:

1. Test package in clean environment
2. Share via appropriate channel (email, repository, cloud storage)
3. Keep archived copy for reference

For quality assurance procedures and checklists, see quality-assurance.md. For common sharing scenarios and example scripts, see common-scenarios.md. For handling cleanup mistakes and notebook deprecation, see cleanup-and-deprecation.md.

Summary

Key principles for project sharing:

1. Choose the right level - Match sharing depth to audience needs
2. Copy, don't move - Preserve your working directory
3. Clean thoroughly - Remove debug code, clear outputs
4. Document everything - README + MANIFEST minimum
5. Check sensitivity - Anonymize or exclude as needed
6. Test before sharing - Run in clean environment
7. Package properly - Compress and document contents
8. Work in main directory - After creating sharing package, ALL future work happens in the original project directory, NOT in the sharing folder

Remember: Good sharing practices benefit both collaborators and your future self!

CRITICAL Reminder for Claude

After creating any sharing package:

1. Always return to the main project directory
2. **Never work in shared-*/ directories** - These are read-only snapshots
3. All future edits, analysis, and development happen in the original working directory
4. Sharing folders are for distribution only, not active development

If the user asks to modify files, always check the current directory and ensure you're working in the main project location, not in a sharing package.