Project Scaffold

Initialize and manage research project structure. Creates standardized directories and metadata files that enable seamless handoffs between skills.

Why This Exists

Without scaffolding, every skill asks "Where are your data?" and "Where should I save this?" Project scaffold solves this by:

Creating canonical paths that all skills respect
Tracking project state so coordinator knows what's done
Enabling prerequisite checking ("Can't write findings until analysis is done")
Providing audit trail for reproducibility

Project Types

Scaffold supports three project types with different structures:

Type	Use When	Key Directories
`qualitative`	Interview research, ethnography, content analysis	transcripts, codebook, memos
`quantitative`	Survey, experiments, secondary data analysis	raw_data, scripts, models, figures
`mixed`	Combining qual + quant methods	Both sets of directories

Two Modes

Mode 1: New Project

Creates full directory structure and blank metadata files.

/project-scaffold

Title: Why Activists Leave Social Movements
Type: qualitative

Mode 2: Adopt Existing Project

Maps existing directories to canonical paths without moving files.

/project-scaffold adopt

My data is in analysis/
My literature is in lit-review/pdfs/

Workflow

Step 1: Gather Project Info

Ask for:

Project title (required)
Project type: qualitative, quantitative, or mixed (required)
Research question (recommended)
Mode: new or adopt

Step 1b: Initialize Git

After gathering project info, ensure the project directory is a git repository:

If not already a git repo, run git init
After creating the directory structure (Step 2), make an initial commit: git add -A && git commit -m "project-scaffold: Initial project structure"

All downstream skills use git for version tracking. Do not create version-suffixed copies of files (e.g., -v2, -final, -working). The git history serves as the version trail.

Step 2: Create/Map Structure

Qualitative Projects

project-name/
├── data/
│   ├── raw/              # Original transcripts, fieldnotes
│   └── clean/            # Processed/anonymized versions
├── analysis/
│   ├── codes/            # Codebook, code definitions
│   ├── memos/            # Analytical memos
│   └── outputs/          # Quote database, matrices
├── library/
│   ├── pdfs/             # PDF and EPUB files
│   └── markdown/         # Converted markdown versions
├── literature/
│   ├── database/         # lit-search outputs
│   ├── notes/            # reading-agent outputs
│   └── synthesis/        # lit-synthesis outputs
├── drafts/
│   ├── sections/         # Individual section drafts
│   └── submission/       # Submission-ready manuscripts
├── submissions/          # Final manuscripts
├── references.bib        # Canonical bibliography (used by all skills)
├── project.yaml          # Project configuration
├── progress.yaml         # State tracking
└── README.md             # Project overview

Quantitative Projects

project-name/
├── data/
│   ├── raw/              # Original datasets
│   ├── processed/        # Cleaned/transformed data
│   └── codebooks/        # Variable documentation
├── scripts/
│   ├── cleaning/         # Data preparation scripts
│   ├── analysis/         # Statistical analysis scripts
│   └── visualization/    # Figure generation scripts
├── models/               # Saved model objects, outputs
├── figures/              # Generated plots and visualizations
├── tables/               # Generated tables
├── library/
│   ├── pdfs/             # PDF and EPUB files
│   └── markdown/         # Converted markdown versions
├── literature/
│   ├── database/
│   ├── notes/
│   └── synthesis/
├── drafts/
│   ├── sections/
│   └── revisions/
├── submissions/
├── references.bib
├── project.yaml
├── progress.yaml
└── README.md

Mixed Methods Projects

project-name/
├── data/
│   ├── qualitative/
│   │   ├── raw/          # Transcripts, fieldnotes
│   │   └── clean/
│   └── quantitative/
│       ├── raw/          # Datasets
│       ├── processed/
│       └── codebooks/
├── analysis/
│   ├── qualitative/
│   │   ├── codes/
│   │   ├── memos/
│   │   └── outputs/
│   └── quantitative/
│       ├── scripts/
│       ├── models/
│       └── outputs/
├── figures/
├── tables/
├── library/
│   ├── pdfs/             # PDF and EPUB files
│   └── markdown/         # Converted markdown versions
├── literature/
│   ├── database/
│   ├── notes/
│   └── synthesis/
├── drafts/
│   ├── sections/
│   └── revisions/
├── submissions/
├── references.bib
├── project.yaml
├── progress.yaml
└── README.md

Step 3: Write Metadata Files

Create project.yaml with user input. Schema varies by project type.

Create progress.yaml with type-appropriate phases.

Step 4: Create README

Generate project README with overview and current status.

Step 5: Report

Show user what was created and next steps.

Progress Tracking by Type

Qualitative Phases

status:
  # Literature
  lit_search: not_started
  lit_synthesis: not_started

  # Analysis
  immersion: not_started
  coding: not_started
  interpretation: not_started
  synthesis: not_started

  # Writing
  theory_draft: not_started
  methods_draft: not_started
  findings_draft: not_started
  bookends_draft: not_started

  # Quality
  verification: not_started
  revision: not_started

Quantitative Phases

status:
  # Literature
  lit_search: not_started
  lit_synthesis: not_started

  # Data
  data_acquisition: not_started
  data_cleaning: not_started

  # Analysis
  eda: not_started              # Exploratory data analysis
  modeling: not_started
  robustness: not_started       # Sensitivity analyses

  # Outputs
  visualization: not_started
  tables: not_started

  # Writing
  theory_draft: not_started
  methods_draft: not_started
  results_draft: not_started
  bookends_draft: not_started

  # Quality
  replication_check: not_started
  revision: not_started

Mixed Methods Phases

status:
  # Literature
  lit_search: not_started
  lit_synthesis: not_started

  # Qualitative strand
  qual_immersion: not_started
  qual_coding: not_started
  qual_interpretation: not_started

  # Quantitative strand
  quant_data_prep: not_started
  quant_eda: not_started
  quant_modeling: not_started

  # Integration
  integration: not_started      # Combining qual + quant

  # Writing
  theory_draft: not_started
  methods_draft: not_started
  findings_draft: not_started
  bookends_draft: not_started

  # Quality
  verification: not_started
  revision: not_started

Progress Dashboard

When invoked with status argument, generate a dashboard:

/project-scaffold status

Dashboard output (qualitative example):

# Project: Why Activists Leave Social Movements
**Type:** Qualitative | **Phase:** Analysis

## Artifacts Found
- [x] Transcripts: 24 files in data/raw/
- [x] Codebook: analysis/codes/codebook.md
- [x] Quote database: analysis/outputs/quote-database.md
- [ ] Theory draft: not found
- [ ] Methods draft: not found

## Status
| Phase | Status |
|-------|--------|
| Literature Search | done |
| Literature Synthesis | done |
| Coding | done |
| Interpretation | in_progress |
| Theory Draft | not_started |

## Suggested Next Steps
1. Complete interpretation - you have codes but no analytical memos
2. Start theory draft with /argument-builder

## Quality Checks Pending
- [ ] Quote verification
- [ ] Bibliography completeness

Dashboard output (quantitative example):

# Project: Income Inequality and Health Outcomes
**Type:** Quantitative | **Phase:** Analysis

## Artifacts Found
- [x] Raw data: data/raw/nhis_2020.dta
- [x] Cleaned data: data/processed/analysis_sample.csv
- [x] Cleaning script: scripts/cleaning/01_prep_data.R
- [x] Analysis script: scripts/analysis/02_main_models.R
- [ ] Figures: 0 files in figures/
- [ ] Tables: 0 files in tables/

## Status
| Phase | Status |
|-------|--------|
| Data Acquisition | done |
| Data Cleaning | done |
| EDA | done |
| Modeling | in_progress |
| Robustness | not_started |
| Visualization | not_started |

## Suggested Next Steps
1. Complete main models
2. Run robustness checks
3. Generate figures and tables

## Quality Checks Pending
- [ ] Replication check
- [ ] Code review

Integration with Other Skills

research-coordinator

On session start:

Check for project.yaml - if missing, invoke project-scaffold
Read progress.yaml and generate dashboard
Route to appropriate skills based on project type

All other skills

Skills read project.yaml for paths:

import yaml
with open("project.yaml") as f:
    config = yaml.safe_load(f)

project_type = config["type"]
if project_type == "qualitative":
    data_path = config["paths"]["transcripts"]
elif project_type == "quantitative":
    data_path = config["paths"]["raw_data"]

# Access library paths (same for all project types)
bib_path = config["library"]["references_bib"]
pdf_dir = config["library"]["pdfs"]
md_dir = config["library"]["markdown"]

Skills update progress.yaml when done:

progress["status"]["coding"] = "done"
progress["artifacts"]["codebook"] = "analysis/codes/codebook.md"
progress["last_updated"] = datetime.now().isoformat()

Commands

Command	What It Does
`/project-scaffold`	Create new project structure
`/project-scaffold adopt`	Map existing project to schema
`/project-scaffold status`	Generate progress dashboard
`/project-scaffold update`	Rescan filesystem and update progress.yaml

Filesystem Scanning by Type

Qualitative Artifacts

Artifact	Detection
Transcripts	`.txt`, `.md`, `*.docx` in `paths.transcripts`
Codebook	`codebook.md` or `codebook.yaml` in `paths.codebook`
Quote database	`quote-database.md` or `quotes.json` in `paths.quotes`
Memos	`*.md` files in `paths.memos`

Quantitative Artifacts

Artifact	Detection
Raw data	`.csv`, `.dta`, `.sav`, `.xlsx` in `paths.raw_data`
Processed data	`.csv`, `.parquet`, `*.rds` in `paths.processed`
Scripts	`.R`, `.py`, `*.do` in `paths.scripts`
Models	`.rds`, `.pkl`, `*.joblib` in `paths.models`
Figures	`.png`, `.pdf`, `*.svg` in `paths.figures`
Tables	`.tex`, `.html`, `*.csv` in `paths.tables`

Schema Versioning

Include schema version for forward compatibility:

# project.yaml
schema_version: 1
type: qualitative  # or quantitative, mixed

If skills encounter newer schema, warn but continue.

Starting the Scaffold

When user invokes:

Check if project.yaml exists
- If yes and no adopt/status flag: "Project already scaffolded. Use /project-scaffold status for dashboard or /project-scaffold adopt to remap paths."
Ask project type:
"What kind of project is this?
1. Qualitative - Interviews, ethnography, content analysis
2. Quantitative - Survey, experiments, secondary data
3. Mixed methods - Combining qual + quant approaches"
Ask for basics:

"Project title?" "Research question (one sentence)?" "Target journals? (optional)"
For adopt mode, ask path questions based on type.
Create structure and files
Report:
"Project scaffolded as [type]. Created:
- [N] directories
- project.yaml (edit to add details)
- progress.yaml (auto-managed by skills)
- README.md
Next steps for [type] project: [type-specific guidance]"