Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Refactor GRACE 4 governed code while keeping .grace graph, verification, change scopes, and file-local anchors synchronized.
.claude/skills/osovv-grace-refactor/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -66% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -14% | 0% |
<skill> <scope> Use this for rename, move, split, merge, extraction, or boundary cleanup work. Greenfield feature work should go through grace-spec, grace-plan, and grace-execute. </scope>
<must_do>
M-*, DF-*, and V-M-* anchors through .grace/graph/index.xml and .grace/verification/index.xml..grace/changes for scope overlap before editing.LINKS:, module contracts, function contracts, and semantic blocks.grace lint --path <project-root> when available.</must_do> </skill>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,283 | 11,101 | +238% | 1 | 1 | 0% | 223 | 566 | +154% | 0 | 0 | — |
case-02 | fail→fail | 10,945 | 7,195 | -34% | 1 | 1 | 0% | 2,144 | 532 | -75% | 0 | 0 | — |
case-03 | fail→fail | 4,402 | 3,228 | -27% | 1 | 1 | 0% | 287 | 385 | +34% | 0 | 0 | — |
case-08 | fail→pass | 11,834 | 3,119 | -74% | 1 | 1 | 0% | 2,019 | 752 | -63% | 0 | 0 | — |
case-04 | fail→pass | 12,104 | 2,636 | -78% | 1 | 1 | 0% | 1,939 | 655 | -66% | 0 | 0 | — |
case-05 | fail→pass | 8,800 | 2,795 | -68% | 1 | 1 | 0% | 1,498 | 681 | -55% | 0 | 0 | — |
case-06 | fail→pass | 10,246 | 2,693 | -74% | 1 | 1 | 0% | 1,792 | 652 | -64% | 0 | 0 | — |
case-07 | fail→fail | 7,731 | 2,876 | -63% | 1 | 1 | 0% | 1,336 | 665 | -50% | 0 | 0 | — |
case-09 | fail→pass | 9,142 | 5,893 | -36% | 1 | 1 | 0% | 1,608 | 1,383 | -14% | 0 | 0 | — |
case-10 | fail→fail | 10,006 | 4,306 | -57% | 1 | 1 | 0% | 1,712 | 988 | -42% | 0 | 0 | — |
case-11 | pass→pass | 8,065 | 5,252 | -35% | 1 | 1 | 0% | 1,436 | 1,216 | -15% | 0 | 0 | — |
case-12 | pass→pass | 11,561 | 4,409 | -62% | 1 | 1 | 0% | 1,907 | 1,037 | -46% | 0 | 0 | — |
case-13 | fail→pass | 8,491 | 6,012 | -29% | 1 | 1 | 0% | 1,523 | 1,387 | -9% | 0 | 0 | — |
case-14 | pass→pass | 10,791 | 4,770 | -56% | 1 | 1 | 0% | 1,798 | 1,061 | -41% | 0 | 0 | — |
case-15 | pass→pass | 8,069 | 4,087 | -49% | 1 | 1 | 0% | 1,363 | 903 | -34% | 0 | 0 | — |
case-16 | fail→fail | 9,196 | 4,675 | -49% | 1 | 1 | 0% | 1,615 | 504 | -69% | 0 | 0 | — |
case-17 | pass→pass | 9,859 | 3,881 | -61% | 1 | 1 | 0% | 1,601 | 736 | -54% | 0 | 0 | — |
case-18 | fail→pass | 6,440 | 2,073 | -68% | 1 | 1 | 0% | 1,260 | 541 | -57% | 0 | 0 | — |
case-19 | fail→pass | 11,829 | 5,694 | -52% | 1 | 1 | 0% | 2,149 | 1,176 | -45% | 0 | 0 | — |
case-20 | fail→fail | 11,161 | 1,915 | -83% | 1 | 1 | 0% | 1,824 | 420 | -77% | 0 | 0 | — |
case-21 | fail→pass | 8,896 | 4,052 | -54% | 1 | 1 | 0% | 1,483 | 880 | -41% | 0 | 0 | — |
case-22 | fail→pass | 20,603 | 4,871 | -76% | 1 | 1 | 0% | 2,003 | 1,069 | -47% | 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 18 counted toward the lift figure. The other 4 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 +45 percentage points is the difference between those two pass rates over the 18 comparable cases.
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.