Internal Study Hub · Otter Group
Tutorial Hub
Self-paced study modules for the team — build, share and reuse the skills we use at Otter Group. Every module runs on the same method described below; pick a module, follow its guides, and get hands-on.
Module 01 — study path
Follow the guides in order — each one builds on the one before. Everything is copy-paste; no coding needed. Total: about one hour, once.
Step 1 · ~30 min
Setup Guide
Install VS Code + Claude Code + DeepSeek on Windows or macOS, incl. troubleshooting.
Step 2 · ~10 min
Skills Installation
Install the gstack toolkit and karpathy-guidelines, create your workspace brief and your first agent.
Step 3 · ~15 min
GitHub Collaboration
Create your GitHub account, connect your laptop, and join the team's private repository.
Step 4 · copy-paste
Sample Finance Tasks
12 day-to-day tasks — cash position, reconciliations, GST, AR aging, budgets and more.
Step 5 · procedure
Consolidated FS
6-stage procedure: subsidiary summaries, eliminations, template, comments & QA.
Modules
Every module below works the same way. Read the general method and the brief user manual once — then start with Module 01 and go one by one.
The general method — VS Code + Claude Code + DeepSeek
Every module uses the same AI assistant setup on your own laptop: VS Code is the editor you open files in, Claude Code is the AI coding agent that runs inside it, and DeepSeek is the AI model in the cloud, reached through a company-issued API key. You describe the task in plain English — Claude Code reads the files in your task folder, does the work step by step, and shows you what it changed. You review and sign off before trusting anything.
One folder per task: copy the files you need into it, open that folder in VS Code, and start Claude Code in its terminal. The rest is prompts — and every module provides the prompts for you.
Brief user manual
- Open VS Code — Terminal → New Terminal (or
Ctrl+`on Windows /Cmd+`on Mac). - Type
claudeand press Enter — the AI assistant starts, working in the folder you opened. - Copy the prompt from the guide, paste it at the
❯prompt, and press Enter. - Watch it work — it reads your files, asks questions if anything is unclear, and makes the changes.
- Review every change before accepting — the AI drafts, you decide. If something fails, paste the error back and it fixes itself.
That's the whole manual. Each module below adds its own guides, prompts and checklists on top of this.
Finance AI Assistant
AVAILABLE NOW Windows 11 & macOS · No coding needed · Company-issued API key required
Put the AI assistant on your own laptop — the same setup used by the GM. Install it, add the skills and agents the finance team uses, connect to the team's GitHub workspace, then use ready-made prompts for your day-to-day finance work and follow the working procedure for the group's consolidated financial report (parent + 3 subsidiaries).
Part 1
Setup Guide
~30 min install for Windows & macOS, incl. troubleshooting table.
Part 2
Skills Installation
~10 min — gstack toolkit, karpathy-guidelines, workspace brief, first agent.
Part 3
GitHub Collaboration
~15 min — account, laptop login, joining the team's private repository.
Part 4
Sample Finance Tasks
12 day-to-day tasks — cash position, reconciliations, GST, AR aging, budgets and more.
Task Sample
Consolidated FS
6-stage procedure: subsidiary summaries, eliminations, template, comments & QA.
Data Analysis with AI
COMING SOON
Turn Excel sheets and CSVs into insight — variance analysis, cost breakdowns, KPI summaries with copy-paste prompts.
Reporting Automation
COMING SOON
Scheduled reports and dashboards — let the machine do the monthly grind: extracts, formatting, and distribution.
Operations & Inventory SOPs
COMING SOON
Standard operating procedures for our outlets and warehouse — searchable, bilingual, always current.
Sales & Marketing Analytics
COMING SOON
Sales dashboards, outlet performance, campaign reviews — the charts behind the numbers.
Module 01 — Finance AI Assistant: how it fits together
You work in VS Code on your own laptop. Claude Code (the AI coding agent) runs inside it and talks to DeepSeek's cloud API using a company-issued key. Everything below the line is covered by one of the guides in Module 01.
Think of the whole setup as a new employee joining your team — it has four parts, and each part has its own job:
| Part | What it is | Simple analogy | Where you meet it |
|---|---|---|---|
| Model (DeepSeek) | The intelligence — the part that actually thinks and answers. | The brain | You never touch it directly — it's a cloud service you configured once with the DeepSeek key. |
| Agent (Claude Code) | The assistant you talk to — reads your files, writes documents, runs commands, asks you questions. | The employee | You type to it in the VS Code terminal; it does the work step by step and shows you everything. |
| Harness (the Claude Code app) | The app that wraps the assistant — permissions, safety checks, session memory, the record of everything you do. | The office / workbench | Every "approve this command?" prompt is the harness working. It also enforces the rules in your CLAUDE.md. |
| Skills (gstack, karpathy-guidelines, …) | Extra toolkits installed into the assistant — like /make-pdf or /context-save. |
The training manuals on the shelf | Installed once in the Skills Installation guide, then called by typing /skill-name. |
In one sentence: the harness is the office, the agent is the employee, the model is the brain, and the skills are the manuals on the shelf.
Where the Overview fits: this page is the hub of the study path. From here you follow the five guides in order — the Setup Guide builds the office (harness + editor), the Skills Installation guide fills the shelf (skills + your first agent), the GitHub guide gives you the shared workspace, and the task guides put the employee to work on real finance jobs. The golden rule stays the same: the AI drafts — you review, verify, and approve before anything is shared.
The local pipeline (top) and where each guide in Module 01 fits (bottom).
Before you start
Three things to have ready — and one rule to keep in mind — before you begin.
- Get a company-issued DeepSeek API key from Finance IT before you begin — you will need it during the Setup Guide.
- Use only company-issued keys. Personal keys, shared logins and trial accounts are not allowed.
- Never paste financial data into free public AI tools. All company financials stay inside the approved setup.