Green — one phase, manually driven

This is a manual phase control, not a separate method. It runs step 3 of the predictive cycle and then stops. Use it when you want to steer Red-Green-Refactor yourself instead of invoking the whole exact-coding workflow. The method lives in one place; this skill only selects which part of it to execute now.

What to read

1. .claude/skills/predictive-tdd/SKILL.md — sections Core rules and 3. Reach Green minimally. Those are binding as written.
2. The matching profile under .claude/skills/predictive-tdd/stacks/ for the commands to run.
3. .claude/skills/exact-coding-shared/human-in-the-loop.md for the prediction-mismatch rule.

What to do

Confirm there is an active behavioral Red to satisfy. If the suite is already green, say so and stop — there is nothing for this phase to do; offer red or refactor instead.

Then execute exactly step 3 of the cycle:

• make the smallest production change likely to pass the active test, without anticipating inactive examples — hardcoded values and narrow conditionals are valid intermediate steps,
• predict the focused test's outcome, including expected pass/fail and todo counts where the stack reports them,
• run it and compare prediction with reality.

Where to stop

Stop after the active test is green. Report the change, the prediction, and the actual result. Do not refactor, and do not start the next behavior — even though Green has no default checkpoint in .claude/skills/exact-coding-shared/human-in-the-loop.md. The human invoked a single phase; that invocation is the checkpoint.

On a prediction mismatch, follow Prediction mismatch in the Predictive TDD skill and stop there. Never change production behavior merely to make an unexplained failure disappear.