Pular para o conteúdo principal

Module 3: Agentic Workflows and Coding Assistants

Learning objectives​

By the end of this module, you will be able to:

  • Place Claude Code, GitHub Copilot and Cursor on the same map and explain what each is designed for.
  • Choose between inline completion, chat and agent mode for a given task, and say why.
  • Describe what "context" means to a model and list the five sources of context you control.
  • Write a project rules file that encodes your team's constraints so you stop repeating them.
  • Recognise when an agent should be stopped and the request restated.

Module content​

Chapter 1: The tool landscape​

Three interaction modes - completion, chat, agent - and the three products most teams evaluate in 2026. What each is good at, where it is weak, how to run a fair evaluation, and why "which tool" matters less than "which workflow".

Chapter 2: Context engineering​

The model only sees its context. Rules files (CLAUDE.md, .cursor/rules, copilot-instructions.md), repository conventions, memory, Model Context Protocol servers and the discipline of keeping the context small enough to be useful.

Quiz​

Six questions.

A warning about product names​

This module names products because the differences matter in practice and because engineers ask. It will date faster than the rest of the course. Read it for the shape of the landscape - the three modes, the role of context, the evaluation method - and check the vendors' documentation for the specifics on the day you need them.