Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Apply when sequencing an addition, refactor, or rewrite. Remove dead weight, redundant validators, and stub references first, then build on the simpler base.
.claude/skills/kunanonj-cursor-plugin-pstack-principle-subtract-before-you-add/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-12 | ✓→✓ | ＝ Same ✓ | 14% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | -65% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -38% | 0% |
When evolving a system, remove complexity first, then build. Deletion gives you a simpler base, which makes the next addition smaller and less brittle.
Why: Adding to a complex system compounds complexity. Removing first cuts the surface area, reveals the essential structure, and usually makes the next design obvious. Default to subtraction.
The pattern:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 9,313 | 8,143 | -13% | 1 | 1 | 0% | 1,698 | 1,665 | -2% | 0 | 0 | — |
case-12 | pass→pass | 3,295 | 3,008 | -9% | 1 | 1 | 0% | 659 | 753 | +14% | 0 | 0 | — |
case-02 | pass→pass | 13,988 | 4,874 | -65% | 1 | 1 | 0% | 3,289 | 1,156 | -65% | 0 | 0 | — |
case-03 | pass→pass | 13,097 | 6,121 | -53% | 1 | 1 | 0% | 2,180 | 1,355 | -38% | 0 | 0 | — |
case-04 | pass→pass | 12,247 | 6,294 | -49% | 1 | 1 | 0% | 2,425 | 1,385 | -43% | 0 | 0 | — |
case-05 | pass→pass | 10,889 | 6,019 | -45% | 1 | 1 | 0% | 2,245 | 1,471 | -34% | 0 | 0 | — |
case-06 | pass→pass | 11,837 | 8,048 | -32% | 1 | 1 | 0% | 2,493 | 1,881 | -25% | 0 | 0 | — |
case-07 | pass→pass | 7,872 | 5,635 | -28% | 1 | 1 | 0% | 1,367 | 1,173 | -14% | 0 | 0 | — |
case-08 | pass→pass | 10,314 | 8,317 | -19% | 1 | 1 | 0% | 2,035 | 1,801 | -11% | 0 | 0 | — |
case-09 | fail→pass | 12,832 | 6,362 | -50% | 1 | 1 | 0% | 2,527 | 1,366 | -46% | 0 | 0 | — |
case-10 | pass→pass | 5,357 | 2,603 | -51% | 1 | 1 | 0% | 968 | 627 | -35% | 0 | 0 | — |
case-11 | pass→pass | 9,793 | 3,881 | -60% | 1 | 1 | 0% | 2,011 | 897 | -55% | 0 | 0 | — |
case-13 | pass→pass | 6,843 | 3,883 | -43% | 1 | 1 | 0% | 1,286 | 913 | -29% | 0 | 0 | — |
case-14 | pass→pass | 9,785 | 4,902 | -50% | 1 | 1 | 0% | 2,079 | 1,102 | -47% | 0 | 0 | — |
case-15 | pass→pass | 7,123 | 4,724 | -34% | 1 | 1 | 0% | 1,457 | 1,120 | -23% | 0 | 0 | — |
case-16 | pass→pass | 8,586 | 4,946 | -42% | 1 | 1 | 0% | 1,820 | 1,052 | -42% | 0 | 0 | — |
case-17 | pass→pass | 7,740 | 4,232 | -45% | 1 | 1 | 0% | 1,634 | 1,069 | -35% | 0 | 0 | — |
case-18 | pass→pass | 7,930 | 6,213 | -22% | 1 | 1 | 0% | 1,556 | 1,418 | -9% | 0 | 0 | — |
case-19 | pass→pass | 9,937 | 4,643 | -53% | 1 | 1 | 0% | 2,111 | 1,083 | -49% | 0 | 0 | — |
case-20 | pass→pass | 7,733 | 4,538 | -41% | 1 | 1 | 0% | 1,724 | 1,107 | -36% | 0 | 0 | — |
case-21 | fail→pass | 22,666 | 7,592 | -67% | 1 | 1 | 0% | 1,365 | 1,680 | +23% | 0 | 0 | — |
case-22 | pass→pass | 7,722 | 4,881 | -37% | 1 | 1 | 0% | 1,703 | 1,172 | -31% | 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.
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.