# claude-code-research-starter **Repository Path**: econometric/claude-code-research-starter ## Basic Information - **Project Name**: claude-code-research-starter - **Description**: https://github.com/luischanci/claude-code-research-starter - **Primary Language**: Unknown - **License**: Not specified - **Default Branch**: main - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2026-04-15 - **Last Updated**: 2026-04-15 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # Claude Code Research Starter Kit A reusable Claude Code workflow for academic research projects in economics, econometrics, macroeconomics, and related fields. Designed for researchers who use **Julia**, **R**, **Stata**, and **LaTeX** for estimation, data analysis, and manuscript preparation. **Author:** Luis Chanci --- ## What This Is A ready-to-use folder structure with Claude Code configuration files (skills, agents, rules, hooks, and templates) tailored for **quantitative economics research**. Clone this repo into your new project and customize the placeholders to get a production-grade Claude Code setup immediately. ### What You Get - **Contractor workflow:** You direct, Claude orchestrates autonomously (plan, execute, verify, report) - **16 custom skills:** 12 workflow skills + 4 econometrics reference skills (Julia, R, Stata, causal inference) - **4 specialized agents:** domain reviewer (substantive correctness), proofreader (grammar/notation), R code reviewer (code quality), verifier (compilation/execution) - **14 rules:** quality gates, code conventions (R + Julia), verification protocols, exploration workflows, session logging - **4 hooks:** file protection, desktop notifications, session log reminders, pre-compaction context snapshots - **3 templates:** session logs, quality reports, exploration READMEs - **Quality gate system:** 80 (commit) / 90 (PR) / 95 (excellence) scoring rubrics --- ## Quick Start ### 1. Clone or Copy ```bash # Option A: Fork on GitHub, then clone your fork git clone https://github.com/YOUR_USERNAME/claude-code-research-starter.git my-new-project cd my-new-project rm -rf .git git init # Option B: Copy into existing project cp -r claude-code-research-starter/.claude/ /path/to/your/project/ cp claude-code-research-starter/CLAUDE.md /path/to/your/project/ cp claude-code-research-starter/MEMORY.md /path/to/your/project/ cp -r claude-code-research-starter/templates/ /path/to/your/project/ cp -r claude-code-research-starter/quality_reports/ /path/to/your/project/ cp -r claude-code-research-starter/explorations/ /path/to/your/project/ cp claude-code-research-starter/.gitignore /path/to/your/project/ ``` ### 2. Launch Claude Code and Paste the Starter Prompt Open a terminal in your project directory and run `claude`. Then paste the following, filling in your project details: > I am starting to work on **[PROJECT NAME]** in this repo. **[Describe your project in 2–3 sentences — what you're building, what data you use, what methods or tools (Julia, R, Stata, LaTeX) are involved.]** > > I want our collaboration to be structured, precise, and rigorous. When creating visuals, everything must be polished and publication-ready. > > I've set up the Claude Code Research Starter Kit (from `luischanci/claude-code-research-starter`). The configuration files are already in this repo. Please read them, understand the workflow, and then **update all configuration files to fit my project** — fill in placeholders in `CLAUDE.md`, adjust rules if needed, and propose any customizations specific to my use case. > > After that, use the plan-first workflow for all non-trivial tasks. Once I approve a plan, switch to contractor mode — coordinate everything autonomously and only come back to me when there's ambiguity or a decision to make. > > Enter plan mode and start by adapting the workflow configuration for this project. **What this does:** Claude reads all the configuration files (`CLAUDE.md`, rules, agents, skills), fills in your project name, institution, and preferences, then enters contractor mode — planning, implementing, reviewing, and verifying autonomously. You approve the plan and Claude handles the rest. **Prefer to configure manually?** Skip to step 3 below for detailed instructions on each file to customize. ### 3. Customize CLAUDE.md (Manual Alternative) Open `CLAUDE.md` and fill in the placeholders marked with `[PLACEHOLDER]`: - **Project title, authors, institution** - **Folder structure** (adapt to your project's layout) - **Commands** (your specific compilation/execution commands) - **Model specifications** (your econometric models) - **Key output files** (your specific outputs) ### 4. Customize Domain-Specific Files 1. **`.claude/agents/domain-reviewer.md`** -- Replace the review lenses with your field's concerns (e.g., identification strategy for causal inference, instrument validity for IV, etc.) 2. **`.claude/rules/knowledge-base-template.md`** -- Fill in your notation registry, symbol reference, and model specifications. 3. **`.claude/rules/r-code-conventions.md`** and **`.claude/rules/julia-code-conventions.md`** -- Adjust the visual palette, common pitfalls, and domain-specific conventions for your project. ### 5. Verify Hooks Work The hooks are configured for **Windows (PowerShell)**. If you're on macOS/Linux, replace `.ps1` files with `.sh` equivalents and update `.claude/settings.json` accordingly. See the `hooks/` section below. ### 6. Start Working ```bash claude # Launch Claude Code in your project directory ``` Claude will automatically read `CLAUDE.md` and load the relevant rules, skills, and agents. --- ## Folder Structure ``` your-project/ ├── CLAUDE.md # Project config (Claude reads every session) ├── MEMORY.md # Persistent learnings across sessions ├── .gitignore # Pre-configured for LaTeX/R/Julia/Python ├── .claude/ # Claude Code configuration │ ├── settings.json # Permissions + hooks │ ├── settings.local.json # Local overrides (gitignored) │ ├── WORKFLOW_QUICK_REF.md # One-page cheat sheet │ ├── agents/ # Specialized review agents │ │ ├── domain-reviewer.md # Substantive correctness review │ │ ├── proofreader.md # Grammar/notation/consistency │ │ ├── r-reviewer.md # R code quality review │ │ └── verifier.md # End-to-end verification │ ├── skills/ # Executable workflows (/command) │ │ ├── compile-latex/SKILL.md # LaTeX compilation │ │ ├── proofread/SKILL.md # Manuscript proofreading │ │ ├── review-r/SKILL.md # R code review │ │ ├── review-julia/SKILL.md # Julia code review │ │ ├── review-paper/SKILL.md # Referee-style paper review │ │ ├── validate-bib/SKILL.md # Bibliography validation │ │ ├── devils-advocate/SKILL.md # Challenge assumptions │ │ ├── commit/SKILL.md # Git workflow │ │ ├── lit-review/SKILL.md # Literature search │ │ ├── research-ideation/SKILL.md # Research questions │ │ ├── interview-me/SKILL.md # Research interview │ │ ├── data-analysis/SKILL.md # End-to-end R analysis │ │ ├── econometrics-julia/SKILL.md # Julia econometrics reference │ │ ├── econometrics-r/SKILL.md # R econometrics reference │ │ ├── causal-inference-r/SKILL.md # IV/DiD/RDD in R │ │ └── stata/ # Comprehensive Stata suite │ │ ├── SKILL.md # Main skill + routing table │ │ ├── references/ (30 files) # Stata reference docs │ │ └── packages/ (20 files) # Community package docs │ ├── rules/ # Embedded governance rules │ │ ├── plan-first-workflow.md │ │ ├── orchestrator-protocol.md │ │ ├── orchestrator-research.md │ │ ├── session-logging.md │ │ ├── verification-protocol.md │ │ ├── quality-gates.md │ │ ├── r-code-conventions.md │ │ ├── julia-code-conventions.md │ │ ├── proofreading-protocol.md │ │ ├── replication-protocol.md │ │ ├── exploration-fast-track.md │ │ ├── exploration-folder-protocol.md │ │ ├── pdf-processing.md │ │ └── knowledge-base-template.md │ └── hooks/ # Automation hooks │ ├── log-reminder.py │ ├── notify.ps1 (or notify.sh) │ ├── protect-files.ps1 (or .sh) │ └── pre-compact.ps1 (or .sh) ├── templates/ # Reference templates │ ├── session-log.md │ ├── quality-report.md │ └── exploration-readme.md ├── quality_reports/ # Plans, logs, merge reports │ ├── plans/ │ ├── session_logs/ │ └── merges/ ├── explorations/ # Research sandbox (60/100 threshold) │ └── ARCHIVE/ ├── scripts/ # Utility scripts └── master_supporting_docs/ # Reference papers └── supporting_papers/ ``` --- ## Skills Reference ### Workflow Skills | Command | What It Does | |---------|-------------| | `/compile-latex [file]` | 3-pass pdflatex + bibtex compilation | | `/proofread [file]` | Grammar, typos, notation consistency review | | `/review-r [file]` | R code quality and reproducibility review | | `/review-julia [file]` | Julia code quality and performance review | | `/review-paper [file]` | Comprehensive referee-style manuscript review | | `/validate-bib` | Cross-reference all citations against .bib | | `/devils-advocate [topic]` | Challenge design decisions and assumptions | | `/commit [msg]` | Stage, commit, PR, merge workflow | | `/lit-review [topic]` | Literature search + synthesis + gaps | | `/research-ideation [topic]` | Generate research questions + strategies | | `/interview-me [topic]` | Interactive research interview | | `/data-analysis [dataset]` | End-to-end R analysis workflow | ### Econometrics Reference Skills | Command | What It Does | |---------|-------------| | `/econometrics-julia` | Julia code patterns for MLE, GMM, simulation-based estimation, Optim.jl, parallel computing | | `/econometrics-r` | R patterns for data.table, fixest, spatial econometrics, causal forests, publication tables | | `/causal-inference-r` | IV, DiD, and RDD in R with proper diagnostics (first-stage F, parallel trends, bandwidth) | | `/stata` | Comprehensive Stata reference: 30 reference docs + 20 package docs (reghdfe, estout, csdid, rdrobust, ivreg2, etc.) | --- ## Quality Gates | Score | Gate | Meaning | |-------|------|---------| | 80 | Commit | Good enough to save | | 90 | PR | Ready for review | | 95 | Excellence | Aspirational | --- ## Workflow Model: Contractor Mode ``` Your instruction | [PLAN] (if non-trivial) --> Show plan --> Your approval | [EXECUTE] Implement, verify, done | [REPORT] Summary + what's ready | Repeat ``` **Claude asks you when:** design forks, code ambiguity, replication edge cases, scope questions. **Claude just executes when:** obvious bug fixes, verification/tolerance checks, documentation/logs, plotting per standards. --- ## Cross-Platform Notes ### Windows (Default) Hooks use PowerShell (`.ps1` files). Works out of the box. ### macOS / Linux 1. Rename hook files from `.ps1` to `.sh` 2. Make them executable: `chmod +x .claude/hooks/*.sh` 3. Update `.claude/settings.json` to call `.sh` instead of `.ps1`: ```json "command": "bash \"$CLAUDE_PROJECT_DIR/.claude/hooks/notify.sh\"" ``` 4. For macOS notifications, use `osascript` instead of Windows toast. --- ## Customization Checklist When starting a new project with this starter kit: - [ ] Fill in `CLAUDE.md` placeholders (project, authors, institution, folder structure) - [ ] Customize `domain-reviewer.md` review lenses for your field - [ ] Update `knowledge-base-template.md` notation registry - [ ] Adjust `r-code-conventions.md` color palette and pitfalls - [ ] Adjust `julia-code-conventions.md` if using Julia (or delete if not) - [ ] Update `verification-protocol.md` with your compilation commands - [ ] Set protected files in `protect-files.ps1` - [ ] Run `claude` and verify hooks work --- ## License MIT -- Use freely for your research projects.