Building Features with Codex: An Agentic Development Workflow

This is the repository for the LinkedIn Learning course Building Features with Codex: An Agentic Development Workflow. The full course is available from LinkedIn Learning.

Course Description

AI coding agents can write code, but shipping reliable features requires a workflow that keeps you in control. In this course, Tiffany Teasley, shows you how to use OpenAI Codex to work through feature requests in an existing application. Learn how to inspect a codebase before making changes, turn requirements into an actionable plan, create a reusable skill for a consistent development process, map acceptance criteria to tests, and build features incrementally. Along the way, see how to reuse the same workflow as the codebase evolves, test features together, diagnose integration issues, and verify your changes without breaking existing functionality.

Learning Objectives

• Inspect an existing codebase with Codex to identify its architecture, dependencies, data flow, tests, and safe extension points before making changes.
• Create and reuse an agent skill that defines a consistent, spec-driven workflow for planning, building, testing, and verifying feature requests.
• Use spec-driven development to translate feature requirements and acceptance criteria into a code-aware implementation and testing plan before coding begins.
• Apply an incremental, test-driven workflow to implement features while preserving existing application behavior.
• Verify integrated features with tests and identify issues that emerge when independently developed functionality works together.

SupportFlow is a small React and TypeScript app for triaging customer support tickets. In this course, you run a feature sprint with OpenAI Codex: map the code, plan the change, build in small steps, test, and review.

Branches

Branch	What it holds
main	Videos 1–5. The starter app.
chapter-6	Videos 6–8. Search and filtering are done. Dashboard metrics aren't built yet.
chapter-9	Video 9. The finished app. Both features work, and all tests pass.

You can use the branch pop-up menu in GitHub to switch branches, or add /tree/BRANCH_NAME to the URL.

When switching branches after making changes to the files, you may get a message like this:

error: Your local changes to the following files would be overwritten by checkout:        [files]
Please commit your changes or stash them before you switch branches.
Aborting

To resolve this issue:

Add changes to git using this command: git add .
Commit changes using this command: git commit -m "some message"

Installing

You need Node 22.9 or newer.

1. Clone this repository into your local machine using the terminal (Mac), CMD (Windows), or a GUI tool like SourceTree.
2. Install and run the app:

npm install
npm run dev

Check it

npm test        # Vitest and Testing Library
npm run lint    # ESLint
npm run build   # type check, then production build

On main, npm test shows 1 passing and 10 failing. That's on purpose. The failing tests in tests/filtering.test.tsx describe Feature 1, and they pass once search and filtering are built.

What's in here

Path	What it's for
PROMPTS.md	Every Codex prompt from the course, ready to copy, with the parts that matter called out
feature-requests/	The two feature requests for the sprint, plus a template for writing your own
.agents/skills/feature-sprint/SKILL.md	The reusable workflow Codex follows for every feature
src/App.tsx	Loads the tickets and passes them to TicketList, which renders each TicketCard
src/components/StatusRing.tsx	Decorative progress panel. It always shows the full data set
src/components/TicketFilters.tsx	Empty stub where Feature 1 goes
src/data/tickets.ts	Ten local tickets and their types. No database or API needed
tests/	The Feature 1 tests. Run them before you change anything

The sprint

1. Feature 1: search and filtering (search-filter.md)
2. Feature 2: dashboard metrics that follow the filters (dashboard-metrics.md)

Instructor

Tiffany Teasley - Data Sistah | Data Scientist and AI Engineer

Tiffany helps aspiring data professionals go from invisible to hired. A full-time data scientist and AI engineer, she spent 20 years as a high school math teacher before making the leap into tech. She failed her first data science interview because no one taught her what companies actually needed. Instead of walking away, she taught herself the skills, built real projects, and came back to land the job. Today through Data Sistah, she's helping the next generation upskill into AI with real-world projects, targeted strategies, and personalized guidance to close the experience gap. A LinkedIn Top Voice in 2024 and 2025, her mission is simple. No developer left behind.

Check out my other courses on LinkedIn Learning.