Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create a detailed refactor plan with tiny commits via user interview, then file it as a GitHub issue. Use when user wants to plan a refactor, create a refactoring RFC, or break a refactor into safe incremental steps.
.claude/skills/fradser-request-refactor-plan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-03 | ✓→✗ | ▼ Worse | 32% | 0% |
Go through the steps below. Skip steps not considered necessary.
<refactor-plan-template>
The problem that the developer is facing, from the developer's perspective.
The solution to the problem, from the developer's perspective.
A LONG, detailed implementation plan. Write the plan in plain English, breaking down the implementation into the tiniest commits possible. Each commit should leave the codebase in a working state.
A list of implementation decisions that were made. This can include:
Do NOT include specific file paths or code snippets. They may end up being outdated very quickly.
A list of testing decisions that were made. Include:
A description of the things that are out of scope for this refactor.
Any further notes about the refactor.
</refactor-plan-template>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 9,949 | 9,760 | -2% | 1 | 1 | 0% | 1,500 | 1,114 | -26% | 0 | 0 | — |
case-02 | fail→pass | 15,296 | 3,344 | -78% | 1 | 1 | 0% | 2,228 | 1,060 | -52% | 0 | 0 | — |
case-03 | pass→fail | 6,796 | 6,650 | -2% | 1 | 1 | 0% | 1,215 | 1,598 | +32% | 0 | 0 | — |
case-04 | pass→fail | 19,077 | 5,603 | -71% | 1 | 1 | 0% | 3,458 | 848 | -75% | 0 | 0 | — |
case-05 | fail→fail | 6,934 | 2,778 | -60% | 1 | 1 | 0% | 1,064 | 956 | -10% | 0 | 0 | — |
case-06 | pass→pass | 15,118 | 8,268 | -45% | 1 | 1 | 0% | 2,248 | 1,742 | -23% | 0 | 0 | — |
case-07 | fail→pass | 14,924 | 6,393 | -57% | 1 | 1 | 0% | 2,439 | 1,556 | -36% | 0 | 0 | — |
case-08 | fail→pass | 12,768 | 2,908 | -77% | 1 | 1 | 0% | 1,893 | 933 | -51% | 0 | 0 | — |
case-09 | pass→fail | 11,861 | 3,575 | -70% | 1 | 1 | 0% | 1,938 | 824 | -57% | 0 | 0 | — |
case-10 | fail→fail | 7,947 | 3,446 | -57% | 1 | 1 | 0% | 1,322 | 1,050 | -21% | 0 | 0 | — |
case-11 | pass→pass | 13,872 | 5,883 | -58% | 1 | 1 | 0% | 2,119 | 1,437 | -32% | 0 | 0 | — |
case-12 | fail→fail | 8,368 | 3,837 | -54% | 1 | 1 | 0% | 1,234 | 1,191 | -3% | 0 | 0 | — |
case-13 | pass→fail | 14,324 | 2,557 | -82% | 1 | 1 | 0% | 2,058 | 951 | -54% | 0 | 0 | — |
case-14 | pass→fail | 16,883 | 5,931 | -65% | 1 | 1 | 0% | 2,364 | 1,487 | -37% | 0 | 0 | — |
case-15 | fail→fail | 10,162 | 2,831 | -72% | 1 | 1 | 0% | 1,251 | 886 | -29% | 0 | 0 | — |
case-16 | pass→pass | 14,515 | 13,900 | -4% | 1 | 1 | 0% | 2,225 | 2,647 | +19% | 0 | 0 | — |
case-17 | fail→fail | 14,120 | 8,951 | -37% | 1 | 1 | 0% | 2,296 | 1,896 | -17% | 0 | 0 | — |
case-18 | fail→fail | 14,415 | 5,744 | -60% | 1 | 1 | 0% | 2,224 | 1,515 | -32% | 0 | 0 | — |
case-19 | pass→fail | 9,351 | 2,434 | -74% | 1 | 1 | 0% | 1,626 | 846 | -48% | 0 | 0 | — |
case-20 | pass→pass | 13,020 | 11,265 | -13% | 1 | 1 | 0% | 2,199 | 1,701 | -23% | 0 | 0 | — |
case-21 | pass→fail | 7,431 | 3,659 | -51% | 1 | 1 | 0% | 1,227 | 1,205 | -2% | 0 | 0 | — |
case-22 | pass→pass | 10,667 | 4,574 | -57% | 1 | 1 | 0% | 1,688 | 1,234 | -27% | 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 20 counted toward the lift figure. The other 2 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 -14 percentage points is the difference between those two pass rates over the 20 comparable cases. 7 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.