# CT AI Developer Launch Kit — shared project guide

## Read this first
This kit uses ordinary Markdown, Python, and JSON files. The included offline example works independently of a coding assistant. You may read and edit these files with Codex, Claude, or another assistant that supports local project files. Compatibility with every assistant has not been tested.

## Project goal
Understand and run the offline workflow, explain each validation step, then improve it with a test. This edition is an executable foundation; the full three-project curriculum and deployed application are still planned.

## Working rules
Use fictional sample data. Keep credentials out of files and conversations. Run the existing tests before and after changing behavior. Explain changes in plain language. Do not connect accounts, install packages, deploy, or enable financial execution as part of this exercise. Stop and explain when a task requires those capabilities.

## First exercise
Read START-HERE.md and workflow_demo.py. Run the example and tests. Explain what the program does and what it does not do. Add a test for a blank question. Then implement a clear validation rule only if the existing validation does not already cover it.

## Completion check
The learner can explain every field, identify fictional data, run tests, reproduce a failure, and describe why this deterministic example is not yet an autonomous AI agent.

## Portability boundary
Project content is portable; tool configuration is not automatically portable. Model API credentials, permission settings, context limits, and assistant-specific features must be configured in the chosen product. No API adapter or live model integration is included in this edition.