Course Overview
AI coding agents can now plan, edit, test, and review software inside a development environment. This two-day workshop teaches you how to direct that work and how to check the result.
It's for experienced software developers. You'll build three projects with Claude Code or Codex: a small CRUD app, an agent that operates a simulation, and a team AI app. Across those projects you'll practise repo instructions, plan mode, tools and MCP, skills, agent loops, evals, token hygiene, and review gates.
You'll work in teams of 3–4. Each participant builds their own CRUD app and simulation agent; teammates review pull requests and share evaluation work. Each team then builds one AI app together. Setup happens in class. The project headings show the learning sequence; they are not a day-by-day timetable.
Lesson Outcomes
By the end of this course, you will be able to:
Direct an agent from a product brief. Write a short PRD, use plan mode, keep standing instructions in
CLAUDE.mdorAGENTS.md, and check the result against the brief.Choose the right kind of instruction. Decide what belongs in a skill, what belongs in an MCP or tool, what belongs in repo instructions, and what should stay as a one-off prompt.
Build an agent loop you can inspect. Give the agent state, typed tools, and hard limits; keep memory outside the chat; keep deterministic work in code; and cap how long the loop can run.
Evaluate behaviour, cost, and quality. Write evals, use subagents as independent checkers, compare models on the same evidence, watch token use and prompt caching, and add quality gates before you merge.
Review and ship with a team. Use worktrees, pull requests, and a second reviewer on agent-written code. Apply the same techniques to a small team app that does not need an agent loop.
Curriculum
Setup
Form a team and prepare Claude Code or Codex, GitHub, and the course tools. You'll also get a short briefing on how coding agents work inside a development environment.
Project 1: CRUD app
Write a one-page PRD and have the agent build a small app from it.
Project 2: Simulation agent
Play a simulation, then build an agent that operates it.
Practise tools, MCP, skills, loops, and persistent state.
Write evals, compare models, and add quality gates.
Project 3: Team AI app
Build a small team app using at least three course techniques.
Presentations and conclusion
Explore Claude Code on the web, present the AI app, and plan what comes next. Each team explains the problem, the input-to-output path, the three techniques they chose, and one item of evidence.
Details
Dates and Times: 26-27 Aug 2026 (2 full-day sessions, 9:00 to 17:00).
Location: CT Hub 2, Lavender.
Requirements: Bring your own laptop. Modern Mac, Windows, or Linux laptops are supported. Participants should be comfortable using a terminal, Git, GitHub, and at least one AI coding tool such as Claude Code or Codex.
Prerequisites: Experienced software developers. Comfortable debugging, reviewing pull requests, and working in a team.
Tools and accounts: The course uses Claude Code or Codex as the main coding tool, with a second tool for comparison. The course setup provides six months of access to the selected 5x plan, using either Claude Code or Codex depending on final setup.
Fees (excl. GST): S$1,710 per participant. Minimum 10 participants to start.
All quoted prices are in SGD and are before 9% GST. Invoicing terms available.