Build, Green, QA: every change earns its place.

The development workflow has three phases, each with its own gate. Build: the code must compile. Green: every test must pass. QA: evidence is recorded for each task.

In Build, each task's worker starts from this brief and writes the code. Its only gate is that the code compiles on main.

When a worker reports, the Dark Lord checks the evidence. Here it finds a passing test summary from a Ritual, and refuses to promote on it. It starts the Test Warden instead, because the laws ask for the Warden's own receipt.

The Test Warden runs the project's full test command once for the whole wave, and reads the result itself.

Its receipt lands on the task: what ran, three tests passed, nothing skipped.

The Dark Lord reads the receipt, sees every test passed, and moves both tasks to New commits to QA.

It plans QA for both tasks and waits for your go-ahead. Home tells you plainly: the Dark Lord waits for you.

After our yes, a QA agent starts on each task. It runs the real app, checks every acceptance criterion, and writes a coverage record.

The Dark Lord reads the record itself: six criteria and two supporting checks, each with how it was checked and the result. It moves the task to QA'd items, and says only you can complete it. The other task's tester is still working.

Open the task and the coverage record is right there, with its limits stated plainly: no separate test machine was available, so QA ran locally.

Row five checks the one function the Test Warden flagged as having no test of its own.

Satisfied, we mark it Done, and it moves to Conquered. Only a person completes a task.