Track A — Cloud Only (No Install)

Step 1 — Create a free Anthropic account

# Open in your browser:
https://claude.ai

Step 2 — Create a Project

Click Projects → New Project. Give it a name (e.g. Workshop). Upload any PDF or text file as context.

Step 3 — You're ready

Follow Demo 5 live (academic homepage from CV). For the /skill demos, watch along — you can run them yourself with Track B later.

Track B — Claude Desktop App + GitHub

Step 1 — Install Claude Desktop

Download from claude.ai/download. Sign in with your Anthropic account.

Step 2 — Connect GitHub MCP

In Claude Desktop: Settings → Extensions → GitHub → Authorize. This lets Claude read and push to your repositories.

Step 3 (optional) — Add local file access

Add a filesystem MCP server to give Claude access to a specific folder on your machine — without using the terminal.

Track C — Full CLI (~15 min)

Step 1 — Install Node.js 18+  (node --version to check; if missing: Mac brew install node · Windows winget install OpenJS.NodeJS · Linux nvm install --lts)

Step 2 — Install Claude Code & authenticate

npm install -g @anthropic-ai/claude-code
claude            # Opens browser → sign in with your Anthropic account
claude --version  # Verify install

Optional — also install Codex or Gemini CLI to compare

npm install -g @openai/codex          # OpenAI Codex CLI
npm install -g @google/gemini-cli    # Google Gemini CLI

Data Retention & Privacy

Before vs. Now

The AI Ladder — Where Are You?

Three CLI Tools at a Glance

A · Tools & Agents

• Claude Code has tools: read file, write file, run shell command, search web.
• An agent = model + tools + loop. It calls tools, reads results, decides the next step — without you driving each step.

Under the hood — the ReAct loop (Reason + Act):

• Three interfaces, one model. The CLI and Desktop App (via MCP) can read and run code in your local files. The web interface works in Claude's own sandboxed environment.
• Skills are shareable instruction sets invoked with a slash: /review, /init. Use /resume to continue a previous session. Anyone can author a skill.

B · CLAUDE.md — Giving the Agent Memory

# My Research Project

## Data
- Source: WRDS / Compustat annual, 1990–2023
- Key file: data/compustat_clean.csv

## Conventions
- Python only · Save all tables to results/ · Always print N before/after filters

C · Permission & Plan Modes

D · Models & Effort

E · Context Windows

• The context window is how much the model holds in memory at once — like RAM, not storage.
• Claude Sonnet 4.6: 200K tokens ≈ a full PhD dissertation in one session.
• Each conversation starts fresh. Use CLAUDE.md to carry over project knowledge.
• Run /compact to summarise prior context and continue, or /clear to reset fully — useful when the window fills mid-task.

F · Plan Limits & Usage Quotas

Pulling a Skill from mixtapetools

# 1 · Create the skills directory in your project
mkdir -p .claude/skills

# 2 · Pull the beautiful-deck skill from Scott Cunningham's repo
curl -o .claude/skills/beautiful-deck.md \
  https://raw.githubusercontent.com/scunning1975/mixtapetools/main/skills/beautiful-deck.md

# 3 · Start Claude Code and invoke it
claude
> /beautiful-deck

1 Referee Report with /referee-2

Step 1 — pull the skill once:

git clone https://github.com/scunning1975/mixtapetools \
  .claude/skills/mixtapetools

Step 2 — run against the paper:

claude "I need to review this paper: paper.pdf
/referee-2
After the structured report, also scan the bibliography:
flag any references where the year or journal name looks incorrect,
and list any key papers in this literature that are missing."

2 Plan Your R&R Response with /strategic-revision

Step 1 — pull the skill once:

git clone https://github.com/jusi-aalto/strategic-revision \
  .claude/skills/strategic-revision

Step 2 — feed it the reviewer reports + your manuscript:

claude "Manuscript: paper.pdf
Reviews: editor-and-reviewer-comments.pdf
/strategic-revision
Output the revision plan to output/."

3 Slide Deck from Paper with /beautiful-deck

Step 1 — pull the skill once:

curl -o .claude/skills/beautiful-deck.md \
  https://raw.githubusercontent.com/scunning1975/mixtapetools/main/skills/beautiful-deck.md

Step 2 — generate the conference presentation:

claude "Here is my working paper: paper.pdf
/beautiful-deck
Target audience: finance conference (EFA / AFA style).
12 slides maximum. Include motivation, data, key table, main figure, conclusion."

4 Replicate Stata Analysis in Python & R

Give Claude your Stata code and let it handle the translation:

claude "I have Stata code in analysis.do running factor regressions.
Replicate the full analysis in Python (pandas + statsmodels) and R (tidyverse + fixest):
  (1) Match sample selection and variable construction exactly
  (2) Reproduce all tables — same column order, same significance stars
  (3) Flag any Stata-specific syntax that required a workaround
Then run /beautiful-deck to turn the comparison into presentation slides."

5 Build Your Academic Homepage from CV

claude "Create a clean academic homepage from cv.pdf.
Sections: About, Research (working papers + publications), Teaching, Contact.
Minimal HTML, no frameworks. Output: index.html"

6 Transcribe a Meeting & Produce a Structured Summary

git clone https://github.com/tchew86/transcribe-meeting \
  .claude/skills/transcribe-meeting
claude
> /transcribe-meeting meeting.m4a

The skill produces a structured summary with:

• Key decisions made — numbered list
• Action items: who is responsible, what, by when
• Open questions that still need resolving
• Suggested agenda items for the next meeting

7 Build Interactive Teaching Tools

Single-prompt build — no prior coding knowledge needed:

claude "Build a student mock-exam interface for a financial markets course.
Use Google Apps Script + Google Sheets as the database. Features:
  - One question at a time, multiple choice, 90-second timer
  - Running score + instant feedback after each answer
  - All submissions logged to a Sheet; admin view shows live results
Deploy as a Google Apps Script Web App."

Starter Prompt Templates

I have a dataset called [YOUR_FILE.csv] with columns [describe].
I want to know [research question].
Please:
  (1) describe the data: rows, columns, missing values
  (2) run an appropriate statistical test
  (3) visualise the key result as a chart
  (4) write a 3-sentence interpretation suitable for a paper

I have [N] Excel/CSV files in a folder, each containing [describe structure].
Write a script that [combines / extracts / calculates / formats] them
and saves the result to output.csv.

Troubleshooting

Take-Home Checklist

• Create CLAUDE.md in every research project — run claude init
• Use --plan before any operation that touches shared files
• Ask Claude to explain its own code: "Walk me through what that script does"
• Use Claude.ai Projects for cross-session memory when working in the browser
• Set a monthly spending cap in the Anthropic Console before using an API key