Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when Java implementation work should be guided by Test-Driven Development, including maintaining a test list, choosing the next behavior, writing a failing test first, implementing only enough production code to pass, and refactoring while keeping tests green. This should trigger for requests such as Apply TDD; Use test-driven development; Drive this Java change with tests; Write the failing test first; Red-green-refactor this feature. Part of Plinth Toolkit
.claude/skills/jabrena-054-design-tdd/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -70% | 0% |
Guide Java developers through Test-Driven Development for implementation work. This is an interactive SKILL.
What is covered in this Skill?
Keep implementation work inside the red-green-refactor cycle, with one selected behavior driving each small change.
references/054-design-tdd.md before applying TDD guidanceRead references/054-design-tdd.md, then list candidate behaviors, edge cases, error paths, integration boundaries, and design questions that may need tests. Add newly discovered cases to the list as work proceeds.
Choose the next useful test case: small enough to implement in one cycle, valuable enough to move the design forward, and specific enough to expose the expected observable outcome.
Write or describe the failing unit, integration, acceptance, or characterization test before production code. Use the test to clarify the public interface, API shape, inputs, outputs, errors, or usage of the code.
Implement the smallest functional production code needed to pass the selected test. Avoid speculative branches, abstractions, framework wiring, or cleanup that is not required by the current failing test.
After the selected test passes, refactor new and existing code to improve names, duplication, responsibility placement, type design, and test clarity while keeping the relevant tests green.
Report the selected behavior, failing-test-first signal, code added to pass, refactoring performed, verification results, skipped checks, missing tests, and remaining risks. Return to the test list for the next cycle.
For detailed guidance, examples, and constraints, see references/054-design-tdd.md.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 14,766 | 11,895 | -19% | 1 | 1 | 0% | 2,439 | 2,879 | +18% | 0 | 0 | — |
case-02 | fail→fail | 13,523 | 11,401 | -16% | 1 | 1 | 0% | 2,093 | 2,737 | +31% | 0 | 0 | — |
case-03 | pass→pass | 10,690 | 11,760 | +10% | 1 | 1 | 0% | 1,875 | 2,803 | +49% | 0 | 0 | — |
case-04 | pass→fail | 21,466 | 5,390 | -75% | 1 | 1 | 0% | 3,113 | 947 | -70% | 0 | 0 | — |
case-05 | pass→fail | 10,021 | 13,428 | +34% | 1 | 1 | 0% | 1,697 | 3,059 | +80% | 0 | 0 | — |
case-06 | pass→pass | 13,200 | 10,061 | -24% | 1 | 1 | 0% | 2,137 | 2,436 | +14% | 0 | 0 | — |
case-07 | pass→pass | 12,376 | 7,946 | -36% | 1 | 1 | 0% | 1,869 | 1,895 | +1% | 0 | 0 | — |
case-08 | pass→pass | 10,041 | 5,915 | -41% | 1 | 1 | 0% | 1,399 | 1,543 | +10% | 0 | 0 | — |
case-15 | fail→pass | 14,558 | 10,519 | -28% | 1 | 1 | 0% | 2,024 | 2,203 | +9% | 0 | 0 | — |
case-09 | pass→pass | 9,661 | 6,144 | -36% | 1 | 1 | 0% | 1,478 | 1,671 | +13% | 0 | 0 | — |
case-10 | pass→pass | 13,581 | 7,544 | -44% | 1 | 1 | 0% | 1,799 | 1,737 | -3% | 0 | 0 | — |
case-11 | pass→pass | 11,830 | 8,864 | -25% | 1 | 1 | 0% | 1,772 | 1,979 | +12% | 0 | 0 | — |
case-12 | fail→pass | 12,287 | 9,096 | -26% | 1 | 1 | 0% | 2,083 | 2,148 | +3% | 0 | 0 | — |
case-13 | pass→pass | 12,318 | 8,984 | -27% | 1 | 1 | 0% | 1,895 | 2,068 | +9% | 0 | 0 | — |
case-14 | pass→pass | 11,416 | 7,319 | -36% | 1 | 1 | 0% | 1,563 | 1,702 | +9% | 0 | 0 | — |
case-16 | pass→pass | 9,774 | 3,307 | -66% | 1 | 1 | 0% | 1,408 | 1,210 | -14% | 0 | 0 | — |
case-17 | pass→pass | 13,317 | 8,998 | -32% | 1 | 1 | 0% | 1,971 | 2,096 | +6% | 0 | 0 | — |
case-18 | pass→pass | 16,757 | 7,331 | -56% | 1 | 1 | 0% | 2,211 | 1,653 | -25% | 0 | 0 | — |
case-19 | pass→pass | 13,569 | 13,391 | -1% | 1 | 1 | 0% | 2,023 | 2,806 | +39% | 0 | 0 | — |
case-20 | pass→pass | 10,188 | 6,842 | -33% | 1 | 1 | 0% | 1,569 | 1,742 | +11% | 0 | 0 | — |
case-21 | fail→pass | 15,884 | 9,384 | -41% | 1 | 1 | 0% | 2,540 | 2,142 | -16% | 0 | 0 | — |
case-22 | pass→pass | 17,677 | 16,165 | -9% | 1 | 1 | 0% | 2,646 | 3,020 | +14% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +9 percentage points is the difference between those two pass rates over the 21 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.