Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Wenn es um BF: Kanzleiwebsite in Barrierefreiheit Web Checker geht: ordnet Sachverhalt, Norm, Beweislast, Gegenargumente und nächsten Schritt; liefert ein direkt nutzbares Arbeitsprodukt mit Prüfpunkten, Risiken und nächstem Schritt.
.claude/skills/klotzkette-bf-kanzleiwebsite-konkret/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 81% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 168% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 64% | 0% |
Frage zu Beginn nur ab, was für den naechsten Schritt unverzichtbar ist. Wenn Material vorliegt, mit dem Material arbeiten und nur eine gezielte Rueckfrage stellen.
Der Output muss als verwertbares Arbeitsprodukt aufgebaut sein:
Dieses Fachmodul arbeitet den konkreten Schwerpunkt aus, prüft Aktenlage, Normen, Fristen, Belege und Gegenargumente und erzeugt einen unmittelbar nutzbaren nächsten Schritt.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 43,810 | 37,607 | -14% | 1 | 1 | 0% | 5,084 | 6,547 | +29% | 0 | 0 | — |
case-02 | fail→fail | 32,502 | 36,107 | +11% | 1 | 1 | 0% | 4,441 | 5,373 | +21% | 0 | 0 | — |
case-03 | fail→fail | 34,119 | 40,297 | +18% | 1 | 1 | 0% | 4,482 | 5,649 | +26% | 0 | 0 | — |
case-04 | fail→pass | 20,232 | 43,474 | +115% | 1 | 1 | 0% | 2,439 | 4,425 | +81% | 0 | 0 | — |
case-11 | pass→pass | 25,357 | 34,615 | +37% | 1 | 1 | 0% | 3,574 | 5,222 | +46% | 0 | 0 | — |
case-05 | fail→pass | 47,939 | 55,349 | +15% | 1 | 1 | 0% | 1,296 | 3,469 | +168% | 0 | 0 | — |
case-06 | pass→fail | 22,346 | 23,915 | +7% | 1 | 1 | 0% | 2,765 | 4,022 | +45% | 0 | 0 | — |
case-07 | pass→pass | 40,683 | 23,158 | -43% | 1 | 1 | 0% | 3,077 | 3,955 | +29% | 0 | 0 | — |
case-08 | fail→pass | 21,964 | 22,764 | +4% | 1 | 1 | 0% | 2,702 | 3,764 | +39% | 0 | 0 | — |
case-09 | pass→pass | 32,708 | 30,017 | -8% | 1 | 1 | 0% | 3,703 | 5,126 | +38% | 0 | 0 | — |
case-10 | fail→pass | 21,494 | 29,445 | +37% | 1 | 1 | 0% | 2,790 | 4,584 | +64% | 0 | 0 | — |
case-12 | fail→fail | 27,978 | 24,607 | -12% | 1 | 1 | 0% | 2,939 | 4,105 | +40% | 0 | 0 | — |
case-13 | pass→pass | 24,579 | 34,137 | +39% | 1 | 1 | 0% | 3,180 | 4,370 | +37% | 0 | 0 | — |
case-14 | fail→pass | 26,062 | 22,498 | -14% | 1 | 1 | 0% | 2,931 | 3,497 | +19% | 0 | 0 | — |
case-15 | pass→pass | 33,714 | 36,539 | +8% | 1 | 1 | 0% | 4,418 | 5,329 | +21% | 0 | 0 | — |
case-16 | fail→fail | 23,695 | 28,218 | +19% | 1 | 1 | 0% | 2,816 | 4,553 | +62% | 0 | 0 | — |
case-17 | pass→pass | 27,950 | 35,118 | +26% | 1 | 1 | 0% | 3,699 | 5,236 | +42% | 0 | 0 | — |
case-18 | fail→pass | 19,666 | 32,506 | +65% | 1 | 1 | 0% | 1,875 | 5,799 | +209% | 0 | 0 | — |
case-19 | pass→fail | 26,184 | 27,718 | +6% | 1 | 1 | 0% | 2,851 | 4,793 | +68% | 0 | 0 | — |
case-20 | fail→fail | 22,072 | 36,159 | +64% | 1 | 1 | 0% | 2,548 | 4,860 | +91% | 0 | 0 | — |
case-21 | fail→pass | 21,622 | 23,030 | +7% | 1 | 1 | 0% | 2,749 | 4,076 | +48% | 0 | 0 | — |
case-22 | fail→pass | 27,558 | 30,510 | +11% | 1 | 1 | 0% | 3,381 | 4,909 | +45% | 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 +32 percentage points is the difference between those two pass rates over the 21 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.