Init Project Research

Interview-driven skill that scaffolds a research project directory and integrates it with the user's Task Management system (context library + Notion).

When to Use

Starting a new research paper or project from scratch
When the user says "new project", "set up a project", "init project", "bootstrap project"
After deciding to pursue a new research idea that needs its own folder

Overview

Seven phases, executed in order:

Interview — gather project details via structured questions
Scaffold — create directory structure based on project type
Seed files — populate CLAUDE.md, README.md, .gitignore with interview answers
Overleaf symlink — link paper/ to Overleaf directory
Git init — initialise repo and make first commit
Task Management sync — update context library and create Notion entry
Confirmation — report what was created

Phase 1: Interview

Use AskUserQuestion for structured input. Three rounds to avoid overwhelming.

Pre-Interview: Auto-Detection

Before asking questions, scan the project directory (if it already exists) for metadata:

LaTeX files — extract \title{}, \author{}, \begin{abstract}, \begin{keyword} from .tex files
Markdown files — check for README.md, notes.md with # Title headings
BibTeX files — note .bib presence for later phases
Overleaf symlink — if paper/ is a symlink, follow and scan the target

Present detected values as the first option (marked "Detected from paper") in interview questions. Always allow override. If the directory doesn't exist yet, skip auto-detection.

Round 1 — Core Identity

Project slug — kebab-case identifier (e.g., costly-voice). Folder name on disk is Title Case with spaces (e.g., Costly Voice). Confirm the derived folder name.
Working title — full paper/project title
Authors / collaborators — names and affiliations
Research area — which parent folder under Research Projects/. Scan for existing theme folders and present as options. Also offer "New topic folder" and "Other location".
Target venue — journal, conference, or preprint:
- Journal: Check CABS AJG ranking via .context/resources/venue-rankings.md and the CSV (.context/resources/venue-rankings/abs_ajg_2024.csv). For SJR score, query the Elsevier Serial Title API (see venue-rankings.md for snippet; requires SCOPUS_API_KEY). Flag journals below CABS 4 with alternatives.
- Conference: Check CORE ranking via .context/resources/venue-rankings.md and the CSV (.context/resources/venue-rankings/core_2026.csv). Capture page limit, format, review type, anonymisation, deadlines.
- Preprint: Note the server (arXiv, SSRN) — no ranking check needed.
Deadline — submission deadline if known

Round 2 — Setup Details

Overleaf project name — folder under ~/Library/CloudStorage/YOUR-CLOUD/Apps/Overleaf/. Verify path exists.
LaTeX template — scan Task Management/templates/ for options. Default: Working Paper (templates/latex-wp/). Also offer "None".
Overleaf external sharing link — read-only URL for collaborators
Git repository? — Local git (Recommended) / GitHub remote / No git
Project type — Experimental (code/, data/, output/) / Computational (src/, tests/, experiments/, results/) / Theoretical (minimal) / Mixed

Round 3 — Research Content

Abstract / elevator pitch — 1-2 sentences
Key research questions — up to 3
Methodology overview — one line

Phase 1.5: Handle Existing Files

If the target directory already exists with files:

Scan for existing files (excluding .claude/)
Read documents to understand content
Present a reorganisation plan: keep in place / move to docs/ / move to docs/readings/ / move to paper/ / move to to-sort/ / absorb into seed files
Wait for approval, execute, double-check before deletions
Use interview content from existing docs to reduce Round 3 questions

If the directory doesn't exist, create it and proceed.

Phase 2: Scaffold Directory

Naming Convention

Slug (kebab-case): example-project — citation keys, git refs
Folder name (Title Case with spaces): Example Project — directory on disk

Overleaf Separation (Hard Rule)

paper/ is for LaTeX source files ONLY. No code, data, scripts, or computational artifacts. See .claude/rules/overleaf-separation.md.

Common Core (always created)

<Folder Name>/
├── CLAUDE.md
├── README.md
├── MEMORY.md
├── .gitignore
├── .context/
│   ├── current-focus.md
│   ├── field-calibration.md
│   └── project-recap.md
├── .claude/
│   ├── hooks/
│   │   └── copy-paper-pdf.sh   # PostToolUse hook — copies paper*/main.pdf → *_vcurrent.pdf
│   └── settings.local.json
├── correspondence/
│   └── reviews/           # .gitkeep (see scaffold-details.md for review structure)
├── docs/
│   ├── literature-review/  # .gitkeep
│   ├── readings/           # .gitkeep
│   └── venues/             # .gitkeep (submission/venue material only)
├── log/                   # .gitkeep
├── paper/                 # Symlink → Overleaf (Phase 4) — LaTeX source ONLY
├── reviews/               # .gitkeep (subdirs created on demand by review agents)
└── to-sort/               # .gitkeep

Conditional Structure

Experimental — add: code/python/, code/R/, data/raw/, data/processed/, output/figures/, output/tables/, output/logs/

Computational — add: src/<project-name>/ (with __init__.py), tests/, experiments/configs/, results/, output/logs/, pyproject.toml, .python-version

Theoretical — nothing extra.

Mixed — present elements and ask which to include.

Venues: When a target venue is known, seed docs/venues/<venue-slug>/submission/. For conference venues, also seed a submission checklist. Full venue structure and checklist template: references/scaffold-details.md.

Python Tooling

Always use uv — never bare pip, python, or requirements.txt. For computational projects, init with uv init. For experimental projects, add pyproject.toml when dependencies are first needed.

Implementation

mkdir -p <dir> && touch <dir>/.gitkeep  # Create all directories
mkdir -p .claude/hooks                   # Create hook, chmod +x
mkdir -p .claude/state                   # Machine-specific memory (gitignored)

Phase 3: Seed Files

CLAUDE.md vs README.md

CLAUDE.md — Instructions for Claude: safety rules, folder structure, conventions, symlink paths
README.md — Human-readable overview: title, authors, abstract, status checklist, links

Both overlap on basic metadata but diverge in purpose. Follow the lean-claude-md rule for CLAUDE.md.

Seed File Templates

Full templates: templates/seed-files.md

File	Purpose
`CLAUDE.md`	Claude instructions: overview, venue, RQs, setup, conventions
`README.md`	Human overview: title, authors, abstract, links, status
`.gitignore`	Standard ignores: OS, IDE, data, paper, Python, R, LaTeX
`MEMORY.md`	Knowledge base: notation, estimands, decisions, pitfalls
`.context/current-focus.md`	Initial "just initialised" state
`.context/field-calibration.md`	Per-project domain profile for agents (placeholder template — `/interview-me` Phase 7 populates it)
`.context/project-recap.md`	Research design notes
`.claude/hooks/copy-paper-pdf.sh`	PDF copy hook
`log/YYYY-MM-DD-HHMM-setup.md`	Initial setup log: project name, creation date, scaffold type, next steps
`docs/pipeline-manifest.md`	(Experimental/Computational only) Script status, data files, manuscript figure manifest. Template: `templates/pipeline-manifest.md`
`run_all.sh`	(Experimental/Computational only) Multi-language pipeline executor (Python via uv, R, Stata). Template: `templates/run-all.sh`. `chmod +x` after creation.

Permissions Sync

After writing .claude/settings.local.json (with hook config), merge global permissions into it so the new project starts with full permissions from day one:

Read ~/.claude/settings.json → extract permissions.allow array
Read the newly created .claude/settings.local.json
Compute the union: local_permissions ∪ global_permissions
Write the merged permissions.allow back to .claude/settings.local.json (preserving the hooks key)

# Merge global permissions into the new project's settings
jq -s '.[0].permissions.allow as $global |
  .[1] | .permissions.allow = ((.permissions.allow // []) + $global | unique | sort)' \
  ~/.claude/settings.json .claude/settings.local.json > .claude/settings.local.json.tmp \
  && mv .claude/settings.local.json.tmp .claude/settings.local.json

Also merge the permissions.deny array using the same logic.

Phase 4: Overleaf Symlink & Template

Create symlink, seed template files if selected, ensure .latexmkrc exists. Full commands and steps: references/scaffold-details.md.

Phase 5: Git Init (conditional)

Skip entirely if the user chose "No git" in Round 2.

cd "<project-path>" && git init && git branch -m main && git add . && git commit -m "Initialize project: <working-title>"

If GitHub remote requested: gh repo create "user/<project-name>" --private --source=. --remote=origin --push

If local git only: remind project syncs via Dropbox. Do NOT push unless a remote was explicitly requested.

Phase 6: Task Management Integration

All paths relative to Task Management root.

6a. Update `.context/projects/_index.md`

Add a new row to the "Papers in Progress" table. Stage is typically "Idea" or "Literature Review".

6b. Create `.context/projects/papers/<short-name>.md`

Template in references/scaffold-details.md.

6c. Update `.context/current-focus.md`

Add to Top 3 Active Projects or as an Open Loop. Use targeted Edit — do NOT rewrite the file.

6d. Create Notion Research Pipeline Entry

Database: YOUR-PIPELINE-DATABASE-ID-HERE
Properties: Name (title), Stage, Target Journal, Co-authors, Priority ("Medium")

Save the Notion page URL for the confirmation report.

Phase 7: Confirmation Report

Created research project: <Working Title>

Directory:  <full path>
Structure:  <N> folders, <N> files
Git:        initialised on branch main (<short commit hash>)
GitHub:     <URL or "local-only (Dropbox sync)">
Overleaf:   paper/ → <target path>

Task Management updates:
  - projects/_index.md:           added row
  - projects/papers/<name>.md:    created
  - current-focus.md:             updated
  - Notion Research Pipeline:     created entry (<URL>)

Setup log:  log/<filename>    created

Next steps:
  1. Open Overleaf and set up main.tex
  2. Run /literature to begin literature review
  3. Start drafting in paper/

Error Handling

Overleaf path doesn't exist: Create symlink anyway (resolves when Overleaf syncs). Warn user.
gh CLI not available: Skip GitHub, note in report.
Notion API fails: Skip Notion entry, offer to retry.
Directory already exists: Ask whether to continue or abort.

Cross-References

Skill	Relationship
`/literature`	Run after init to begin literature search
`/project-safety`	Already handled — .gitignore and settings created during init
`/save-context`	Context library entries created during Phase 6
`/session-log`	Offer to create a session log after init completes
`/interview-me`	To develop the research idea before scaffolding

init-project-research