ICCT logo

Internal academy

ICCT Training Library

AI & Agent Workflows

AI Coding Agents & Editors

Cover modern coding agents and editors such as Codex, Cursor, Claude Code, and Gemini Code Assist so students learn when to supervise, when to review, and how to ship safely.

Intermediate 3 credits 50+ guided hours 5 weeks Modules: 3 Projects: 2

Easy Interactive curriculum

Signal Lab for short, easy-to-digest learning

This course breaks big ai & agent workflows ideas into short missions, plain-language definitions, and interactive practice rounds. Cover modern coding agents and editors such as Codex, Cursor, Claude Code, and Gemini Code Assist so students learn when to supervise, when to review, and how to ship safely.

Class missions: 3 10 minutes per topic Lab + quiz + homework + game

Class mission

From assistant to agent

Minutes: 20

Big idea: Understand the shift from inline suggestions to tools that can search, edit, and run tasks. We keep it simple, hands-on, and easy to explain back.

10 min

Codex and cloud agent workflows

Codex means a coding helper that can read project context and suggest or make code changes In this topic we use it for parallel task delegation, worktrees, and agent supervision.

Quick picture: imagine using this in prototype delivery.

Mission: use Codex to practice parallel task delegation, worktrees, and agent supervision.

Codex Cloud Gemini Code Assist Claude Code
10 min

Claude Code and Gemini CLI in the terminal

Claude Code means a terminal coding helper that reads files, edits work, and follows task steps In this topic we use it for command-line agents that read, edit, and run tasks locally.

Quick picture: imagine using this in bug fixing labs.

Mission: use Claude Code to practice command-line agents that read, edit, and run tasks locally.

Claude Code Gemini CLI Claude Gemini

Word wall

Terms made easy

Codex

a coding helper that can read project context and suggest or make code changes

Cloud

computers and services you use over the internet instead of your own machine

Gemini Code Assist

Google coding help that explains code and helps make changes inside tools

Claude Code

a terminal coding helper that reads files, edits work, and follows task steps

Lab mission

Lab Mission

Compare one bug-fix task across a cloud agent and a terminal agent, then document the tradeoffs.

  1. Look at one example from prototype delivery and point to where Codex shows up.
  2. Build one tiny version that uses Codex and Cloud in the right place.
  3. Tell a teacher or partner what you changed, what worked, and what you would improve next.
Homework

Homework: teach Codex in your own words, then create agent playbook with one example, one picture, or one tiny screenshot.

Quick quiz

Check understanding in a friendly way

Game mode

Term Match Sprint

Matches: 0 / 4

Match each word to the right meaning as fast as you can.

Words
Meanings

Pick a word, then pick the matching meaning.

Outcomes

What students should leave with

  • Use Codex, Cursor, Claude Code, and Gemini Code Assist for different build and review tasks.
  • Delegate coding work while preserving human approval, testing, and architecture control.
  • Compare IDE, terminal, and cloud-agent workflows across real project scenarios.

Reference map

Core topics to cover

Agent prompting Context and codebase awareness Multi-file edits Terminal agents IDE copilots Testing loops Human-in-the-loop review

Module sequence

ICCT lesson flow

Module 1

From assistant to agent

Lessons: 2

Understand the shift from inline suggestions to tools that can search, edit, and run tasks.

Codex and cloud agent workflows

parallel task delegation, worktrees, and agent supervision

Claude Code and Gemini CLI in the terminal

command-line agents that read, edit, and run tasks locally

Checkpoint lab

Compare one bug-fix task across a cloud agent and a terminal agent, then document the tradeoffs.

Module 2

IDE-native AI editing

Lessons: 2

Use editors that stay close to code while still supporting natural-language changes.

Cursor for context-aware edits

asking questions, applying edits, and codebase-aware drafting

Gemini Code Assist inside the IDE

chat, completions, explanations, and transformation workflows

Checkpoint lab

Use an AI editor to implement a small feature, then review every change before accepting it.

Module 3

Review, test, and ship

Lessons: 2

Treat AI output like teammate output: useful, but not self-approving.

Prompting for testable changes

clear scope, acceptance criteria, and checklists

Human approval and safe delivery

test runs, code review, rollback thinking, and risk notes

Checkpoint lab

Ship a reviewed feature branch with acceptance notes, test evidence, and a short change summary.

Capstone prompts

Projects students can actually ship

Coding Agent Playbook

Define when your team should use Codex, Cursor, Claude Code, or Gemini Code Assist for different tasks.

Supervised Feature Build

Deliver one feature using AI assistance, but with explicit test evidence and human review checkpoints.