Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Clean-break refactors with migration for the Research Project Template — move logic to src/, split modules, rename APIs with test updates. USE WHEN restructuring code, extracting modules, removing duplication, or migration without behavior change.
.claude/skills/docxology-template-refactoring/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -40% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -62% | 0% |
src/ or infrastructure/; keep orchestrators thin.bashuv run pytest <affected-test-path> -v uv run python -m infrastructure.project.public_scope lint-paths | xargs uv run ruff check uv run python scripts/pipeline/stage_01_test.py --project <project> # when project scope
docs/core/architecture.mduv run python scripts/gates/module_line_count_check.py| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-11 | fail→pass | 6,894 | 2,382 | -65% | 1 | 1 | 0% | 1,174 | 678 | -42% | 0 | 0 | — |
case-01 | fail→fail | 4,560 | 4,740 | +4% | 1 | 1 | 0% | 280 | 655 | +134% | 0 | 0 | — |
case-02 | fail→fail | 5,788 | 4,627 | -20% | 1 | 1 | 0% | 261 | 607 | +133% | 0 | 0 | — |
case-03 | fail→fail | 6,069 | 4,527 | -25% | 1 | 1 | 0% | 271 | 587 | +117% | 0 | 0 | — |
case-04 | pass→pass | 17,921 | 4,635 | -74% | 1 | 1 | 0% | 3,508 | 1,134 | -68% | 0 | 0 | — |
case-05 | pass→fail | 8,124 | 5,602 | -31% | 1 | 1 | 0% | 1,633 | 717 | -56% | 0 | 0 | — |
case-06 | pass→fail | 14,244 | 6,109 | -57% | 1 | 1 | 0% | 2,776 | 522 | -81% | 0 | 0 | — |
case-07 | fail→fail | 7,176 | 3,485 | -51% | 1 | 1 | 0% | 1,289 | 1,024 | -21% | 0 | 0 | — |
case-08 | fail→pass | 10,069 | 4,177 | -59% | 1 | 1 | 0% | 1,763 | 1,051 | -40% | 0 | 0 | — |
case-09 | fail→pass | 8,347 | 1,626 | -81% | 1 | 1 | 0% | 1,490 | 607 | -59% | 0 | 0 | — |
case-10 | pass→pass | 6,907 | 4,002 | -42% | 1 | 1 | 0% | 1,256 | 904 | -28% | 0 | 0 | — |
case-12 | pass→pass | 7,372 | 3,312 | -55% | 1 | 1 | 0% | 1,439 | 840 | -42% | 0 | 0 | — |
case-13 | pass→pass | 12,575 | 5,827 | -54% | 1 | 1 | 0% | 1,961 | 1,450 | -26% | 0 | 0 | — |
case-14 | pass→pass | 3,826 | 1,450 | -62% | 1 | 1 | 0% | 661 | 653 | -1% | 0 | 0 | — |
case-15 | fail→pass | 6,352 | 1,722 | -73% | 1 | 1 | 0% | 1,004 | 661 | -34% | 0 | 0 | — |
case-16 | pass→pass | 3,931 | 2,139 | -46% | 1 | 1 | 0% | 685 | 702 | +2% | 0 | 0 | — |
case-17 | fail→pass | 12,446 | 2,558 | -79% | 1 | 1 | 0% | 2,014 | 758 | -62% | 0 | 0 | — |
case-18 | fail→pass | 7,116 | 2,202 | -69% | 1 | 1 | 0% | 1,205 | 548 | -55% | 0 | 0 | — |
case-19 | pass→pass | 12,252 | 3,870 | -68% | 1 | 1 | 0% | 1,631 | 935 | -43% | 0 | 0 | — |
case-20 | pass→pass | 12,766 | 6,007 | -53% | 1 | 1 | 0% | 1,986 | 1,289 | -35% | 0 | 0 | — |
case-21 | pass→pass | 6,761 | 2,782 | -59% | 1 | 1 | 0% | 1,041 | 665 | -36% | 0 | 0 | — |
case-22 | pass→pass | 4,348 | 2,524 | -42% | 1 | 1 | 0% | 741 | 815 | +10% | 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 17 counted toward the lift figure. The other 5 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 +18 percentage points is the difference between those two pass rates over the 17 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.