Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Ponytail, lazy senior dev mode. Always pick the simplest solution that works.
.claude/skills/kunanonj-cursor-plugin-ponytail-rule-ponytail/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-20 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -80% | 0% |
You are a lazy senior developer. Lazy means efficient, not careless. The best code is the code never written.
Before writing any code, stop at the first rung that holds:
Rules:
ponytail: comment. If the shortcut has a known ceiling (global lock, O(n²) scan, naive heuristic), the comment names the ceiling and the upgrade path.Not lazy about: input validation at trust boundaries, error handling that prevents data loss, security, accessibility, anything explicitly requested. Non-trivial logic leaves ONE runnable check behind, the smallest thing that fails if the logic breaks (an assert-based demo/self-check or one small test file; no frameworks, no fixtures). Trivial one-liners need no test.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | fail→pass | 11,971 | 10,944 | -9% | 1 | 1 | 0% | 2,519 | 2,579 | +2% | 0 | 0 | — |
case-01 | fail→pass | 7,206 | 6,306 | -12% | 1 | 1 | 0% | 1,699 | 1,527 | -10% | 0 | 0 | — |
case-02 | fail→pass | 10,818 | 8,111 | -25% | 1 | 1 | 0% | 2,431 | 2,100 | -14% | 0 | 0 | — |
case-03 | fail→pass | 9,627 | 5,710 | -41% | 1 | 1 | 0% | 1,854 | 1,255 | -32% | 0 | 0 | — |
case-04 | pass→pass | 17,851 | 10,596 | -41% | 1 | 1 | 0% | 3,528 | 2,365 | -33% | 0 | 0 | — |
case-05 | pass→pass | 11,601 | 12,028 | +4% | 1 | 1 | 0% | 2,279 | 2,582 | +13% | 0 | 0 | — |
case-06 | pass→pass | 22,160 | 4,876 | -78% | 1 | 1 | 0% | 4,974 | 1,290 | -74% | 0 | 0 | — |
case-07 | fail→pass | 15,766 | 2,315 | -85% | 1 | 1 | 0% | 3,983 | 779 | -80% | 0 | 0 | — |
case-08 | pass→pass | 7,881 | 2,290 | -71% | 1 | 1 | 0% | 1,629 | 708 | -57% | 0 | 0 | — |
case-09 | pass→pass | 5,222 | 3,619 | -31% | 1 | 1 | 0% | 1,051 | 1,002 | -5% | 0 | 0 | — |
case-10 | pass→pass | 6,625 | 2,445 | -63% | 1 | 1 | 0% | 1,289 | 741 | -43% | 0 | 0 | — |
case-11 | pass→pass | 8,296 | 2,076 | -75% | 1 | 1 | 0% | 1,409 | 686 | -51% | 0 | 0 | — |
case-12 | pass→pass | 6,357 | 3,009 | -53% | 1 | 1 | 0% | 1,180 | 900 | -24% | 0 | 0 | — |
case-13 | pass→pass | 3,552 | 3,283 | -8% | 1 | 1 | 0% | 662 | 928 | +40% | 0 | 0 | — |
case-14 | pass→pass | 4,503 | 3,203 | -29% | 1 | 1 | 0% | 924 | 832 | -10% | 0 | 0 | — |
case-15 | pass→pass | 6,109 | 3,111 | -49% | 1 | 1 | 0% | 1,231 | 925 | -25% | 0 | 0 | — |
case-16 | fail→pass | 10,081 | 2,897 | -71% | 1 | 1 | 0% | 2,443 | 897 | -63% | 0 | 0 | — |
case-17 | pass→pass | 5,309 | 2,989 | -44% | 1 | 1 | 0% | 906 | 876 | -3% | 0 | 0 | — |
case-18 | fail→pass | 8,448 | 5,724 | -32% | 1 | 1 | 0% | 1,694 | 1,411 | -17% | 0 | 0 | — |
case-19 | pass→pass | 3,197 | 3,016 | -6% | 1 | 1 | 0% | 603 | 904 | +50% | 0 | 0 | — |
case-21 | pass→pass | 10,872 | 4,888 | -55% | 1 | 1 | 0% | 1,836 | 1,210 | -34% | 0 | 0 | — |
case-22 | pass→pass | 10,024 | 4,903 | -51% | 1 | 1 | 0% | 1,959 | 1,276 | -35% | 0 | 0 | — |
case-23 | pass→pass | 7,438 | 3,146 | -58% | 1 | 1 | 0% | 1,665 | 891 | -46% | 0 | 0 | — |
case-24 | fail→pass | 12,901 | 7,479 | -42% | 1 | 1 | 0% | 2,875 | 1,688 | -41% | 0 | 0 | — |
case-25 | pass→pass | 5,540 | 4,302 | -22% | 1 | 1 | 0% | 1,020 | 1,053 | +3% | 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. 25 cases were attempted. The headline lift of +32 percentage points is the difference between those two pass rates over the 25 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.