AI Engineering Starter Kit

A production-oriented catalogue for consistent, secure AI-assisted development. The repository is the durable context: people and AI tools read the same version-controlled rules.

Java, Python and C# are end-to-end runnable profiles with templates, skills, verification, and multistage Docker/Compose/Dev Container assets (development toolchain baked from pinned images). All other language profiles are standards-only scaffolds and require a product build/test payload.

Quick start

1. Read AGENTS.md and .ai/project.md.
2. Install Java 21, Maven 3.9+, Docker, and Git.
3. Run git config core.hooksPath hooks.
4. Run ./scripts/verify.sh (or scripts\verify.ps1 on Windows).
5. Run mvn verify inside templates/java to verify the Java reference template.

Non-negotiable gates

• Human ownership of every change and approval.
• No secrets, personal data, client data, or proprietary code in unapproved models.
• Every test contains at least one outcome assertion; interaction verification alone is insufficient.
• Formatting, static analysis, tests, dependency checks, secret scanning, and CI must pass.
• AI output is untrusted until inspected.

See docs/onboarding/GETTING_STARTED.md, .ai/skills.md, and .ai/prompts.md. Presentation handouts: docs/handouts/ (Markdown) and the generated PDFs listed in docs/README.md.

Starting a new repository

Read docs/QUICKSTART.md before using this kit as a project template. It explains what to rename, replace, retain or delete and how to activate backend, native, Node.js, React or Next.js context profiles.

Option A — Swagger UI (recommended)

Requires uv and PowerShell (pwsh or powershell) on PATH.

tools\project-generator-api\scripts\run.ps1

Then open http://127.0.0.1:8090/docs

1. Expand POST /projects → Try it out
2. Fill the form fields (not raw JSON):
   • profile — dropdown (e.g. java)
   • project_name — e.g. billing-api
   • destination — paste a normal Windows path, e.g. C:\Tools\Workspace\Auth (no escaping)
3. Execute — the API runs scripts/create-project.ps1

Stop the server with Ctrl+C in that terminal. Details: tools/project-generator-api/README.md.

Option B — script

Create in a parent folder (project folder is created as <Destination>\<ProjectName>):

scripts\create-project.ps1 -Profile java -ProjectName product-api -Destination C:\Workspace

Or pass the full project path (folder name must match -ProjectName):

scripts\create-project.ps1 -Profile java -ProjectName product-api -Destination C:\Workspace\product-api

Each generated project includes <ProjectName>.md — product-specific AI context to enrich with purpose, scope and working notes. Agents are instructed to read it before changing code.

Runnable profiles: java, python, csharp. Scaffold-only profiles still receive governance and language context, but no application template or stack skills yet.

Reusable development skills

The skills/ directory contains Codex skills for Java (6), Python (4), and C# (4): feature work, tests, review, and security review, plus Java diagnosis and refactor. Every skill loads the repository's Markdown context before working. See .ai/skills.md for installation and invocation.