Module 6: Test-Driven Development

Pillar 05: "Green and clean, no exceptions."

What you need to know first

Test: A small program that runs your program and checks if it does the right thing. Instead of you manually running the program and typing inputs to check, a test does it automatically.

Example:

# Your function
def add_hours(existing, new):
    return existing + new

# A test for that function
def test_add_hours():
    assert add_hours(2.0, 1.5) == 3.5      # normal case
    assert add_hours(0, 5) == 5              # zero case
    assert add_hours(0, 0) == 0              # both zero

The assert statement says "this must be true, and if it's not, the test fails." You run all your tests after every change, and if any fail, you know exactly what broke.

pytest: A Python tool that finds and runs all your test files automatically. You'll install it with pip install pytest and run it with pytest in your terminal.

Test-Driven Development (TDD): Writing the tests before writing the code. This sounds backward, but it forces you to think about what the code should do before thinking about how it does it.

The SOLO idea

In agentic coding, tests are your safety net. The AI will change your code — sometimes correctly, sometimes not. Tests catch the "not" cases automatically. Without tests, you're manually checking every feature after every AI change. With tests, you type pytest and know in 2 seconds if anything broke.

SOLO adds one critical rule: the AI is never allowed to modify a test to make it pass. If a test fails, the AI must fix the code, not the test. This prevents the AI from silently lowering your standards.

Reading

• SOLO blog: Pillar 05
• Open workflows/test.md in VS Code — note the strict prohibition on test hacking
• Open workflows/code.md — note "Truth in Tests" and "Mandatory TDD"

---

Lab 6: Tests as a Safety Net

Estimated AI conversations: 2-3

1. Set up pytest:

   pip install pytest
   

2. Write tests first. Before implementing anything, type in Copilot Chat:

   Based on #file:SPEC.md, I want to write tests for the study tracker. We're using pytest. Write test functions for these scenarios — but DON'T write the actual implementation yet:

   • Logging a session with valid data saves it correctly
   • Logging a session with empty class name is rejected
   • Logging a session with negative hours is rejected
   • Logging a session with non-numeric hours is rejected
   • Listing sessions when no sessions exist shows an empty message
   • Deleting a session that exists removes it
   • Deleting a session that doesn't exist shows an error
   • Summary shows correct total hours per class

3. Run the tests — they should all fail (since the code doesn't exist yet):

   pytest -v
   

   This is correct! Failing tests are the starting point.

4. Now implement. In Copilot Chat, type:

   Follow #file:workflows/code.md — Here is my spec: #file:SPEC.md and my plan: #file:PLAN_v2.md. Implement the study tracker so that all existing tests pass.

5. Run tests again:

   pytest -v
   

   Watch them go green. If any fail, tell the AI: "These tests are failing. Fix the code, not the tests."

6. Test your safety net. Open one of the source files and make a small, intentional change — for example, remove the check that rejects negative hours. Run pytest again. Does a test catch it? If yes, the net works. Undo your change (git checkout -- .).

7. Commit:

   git add .
   git commit -m "Implemented with tests"
   git push
   

The mental model: Tests are a contract between you and the AI. You define what "correct" means. The AI is free to write whatever code it wants, as long as the tests pass. Weak tests = weak code. Rigorous tests = rigorous code.