Claude Code Training
A practical guide for Data Scientists and ML Engineers.
Part 1
Claude Code: Zero to Proficient
Prompt Mode and Prefixes
Claude Code's three operating modes and the four input prefixes — what they do, when to use them, and how to combine them.
What is CLAUDE.md?
CLAUDE.md is the configuration file Claude Code reads automatically every session — permanent project context without re-explaining your setup each time.
Tokens, Context and Pricing
Understand what tokens are, how the context window works, and how to keep your Claude Code usage costs under control.
Basic Commands
Learn the essential Claude Code CLI commands — interactive and non-interactive modes, shortcuts, and key options.
Custom Commands
Create and activate your own slash commands to automate repetitive workflows with Claude Code.
Skills and Meta-Skills
Reusable prompt templates that Claude can invoke automatically — how to create, activate, and write descriptions for automatic invocation.
Agents
Understand what agents are in Claude Code, how they differ from skills, and how to create and use them effectively.
Configuration Sharing
Learn how to share skills, agents, commands, and MCP servers as plugins within your team and the wider community.
MCP Integrations
Extend Claude Code with the Model Context Protocol — connect your databases, APIs, and internal tools.
Token Reduction Strategies
Advanced strategies to reduce token consumption: @ commands, context management, and RTK techniques.
Part 2
Spec-Driven Development
Why Spec-Driven Development?
The specific failure modes that emerge when working with Claude without a spec — and why a written specification fixes them.
Spec-Driven Development Concepts
What a spec looks like, what goes in each section, and when SDD is worth the overhead.
Spec-Driven Frameworks
OpenSpec, spec-kit, BMAD, and custom templates compared — which one fits your context and how to get started.