# Lab 5 — Tests + CI/CD

> **Summary — what this page covers**
> Attendees generate an AI-written test suite for their Day 2 features, then add a GitHub Actions
> workflow that does **both** halves of the outcome: automated **AI code review** on PRs and
> **AI-generated release notes** on tag push — committed. Prioritize the testing half; CI YAML can
> be largely provided as reference.

**Duration: 25 min · Deliverable: AI-generated tests passing + GitHub Actions AI review + release-notes job, committed**

## Part A — Generate tests (≈10 min)

Use Claude Code to generate an xUnit suite for `AgentService` — happy path, a multi-tool run, a tool
error, and a `CancellationToken` path — using **NSubstitute** and the existing patterns in
`BookTracker.Tests`. Then run the **adversarial edge-case prompt** (see
[Section 5](11-section-5-testing-cicd.md)) and add the cases it surfaces. Run `dotnet test` until green.

## Part B — Integration tests with TestContainers (≈stretch)

```bash
dotnet add BookTracker.Tests package Testcontainers.MsSql
dotnet add BookTracker.Tests package Testcontainers.Qdrant
```

Write **one integration test** that exercises the **real SQL Server + Qdrant** stack via
TestContainers — no mocks. Optional if you're short on time.

## Part C — AI code review in GitHub Actions (≈8 min)

Add `.github/workflows/ai-code-review.yml`: on a PR, diff the change and run
`claude --no-interactive --model claude-haiku-4-5` to review for **security / .NET issues / missing
error handling / CLAUDE.md violations**, then post the result as a **PR comment**. Frame this
workflow as your reusable **code-review skill** — the same review you run locally with `/code-review`,
now in CI (the Day 1 steering touchpoint).

## Part D — AI-generated release notes (≈7 min)

Add a release-notes job triggered on **tag push**: collect commits since the previous tag, pipe them
to `claude --no-interactive` (Haiku, to keep CI cheap), and produce notes grouped as **What's New /
Bug Fixes / Breaking Changes / Developer Notes**, attached to the GitHub Release.

```bash
git log ${PREV_TAG}..HEAD --format="%H %s %b" > commits.txt
```

## Checkpoint

- [ ] AI-generated tests pass (`dotnet test`)
- [ ] Edge-case tests added from the adversarial pass
- [ ] CI workflow runs the AI **code review** on a PR
- [ ] Tag push produces AI-generated **release notes**
- [ ] (Stretch) one TestContainers integration test passes
- [ ] Committed to your fork

## Bonus — The Full CI/CD Pipeline

No time pressure. Extend the workflow into a complete pipeline:

- **Build/test gates** — run `dotnet build` and `dotnet test` as required checks before merge.
- **A quality gate** — have the AI review **fail the build on critical findings**, not just comment.
- **Skill-usage logging** — add the `PreToolUse` skill-logging hook from Day 1 so you can see which
  skills your CI actually triggers.