Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Wenn es um Eigentumsvorbehalt in AGB-Recht-Prüfer geht: ordnet Sachverhalt, Norm, Beweislast, Gegenargumente und nächsten Schritt; liefert eine Fristen- und Risikoampel mit Sofortschritten. Auswahlstichwort: Agb Eigentumsvorbehalt B2b B2c; Arbeitsfeld: AGB-Recht-Prüfer.
.claude/skills/klotzkette-agb-eigentumsvorbehalt-b2b-b2c/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-10 | ✓→✗ | ▼ Worse | 24% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 33% | 0% |
> Die gelieferte Ware bleibt bis zur vollständigen Bezahlung sämtlicher Forderungen aus der Geschäftsverbindung Eigentum des Verkäufers (Eigentumsvorbehalt). Der Käufer ist berechtigt, die Vorbehaltsware im ordentlichen Geschäftsgang weiterzuveräußern. Die Forderungen aus dem Weiterverkauf tritt der Käufer hiermit zur Sicherheit in Höhe des Rechnungsendbetrags (einschließlich Umsatzsteuer) an den Verkäufer ab; der Verkäufer nimmt die Abtretung an. Der Käufer bleibt zur Einziehung berechtigt, solange er seinen Zahlungsverpflichtungen nachkommt.
Siehe references/QUELLEN.md, references/PRUEFLOGIK.md und references/KLAUSELFAMILIEN.md.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,868 | 14,513 | +13% | 1 | 1 | 0% | 2,354 | 3,999 | +70% | 0 | 0 | — |
case-02 | pass→pass | 15,096 | 14,606 | -3% | 1 | 1 | 0% | 2,735 | 3,645 | +33% | 0 | 0 | — |
case-03 | fail→pass | 14,999 | 10,191 | -32% | 1 | 1 | 0% | 2,879 | 2,921 | +1% | 0 | 0 | — |
case-04 | pass→pass | 12,600 | 15,403 | +22% | 1 | 1 | 0% | 2,691 | 4,364 | +62% | 0 | 0 | — |
case-05 | fail→fail | 21,149 | 15,911 | -25% | 1 | 1 | 0% | 3,575 | 4,205 | +18% | 0 | 0 | — |
case-06 | pass→pass | 13,321 | 13,623 | +2% | 1 | 1 | 0% | 2,858 | 3,916 | +37% | 0 | 0 | — |
case-07 | pass→pass | 17,713 | 16,121 | -9% | 1 | 1 | 0% | 3,104 | 3,997 | +29% | 0 | 0 | — |
case-08 | fail→fail | 12,128 | 10,618 | -12% | 1 | 1 | 0% | 2,055 | 3,070 | +49% | 0 | 0 | — |
case-09 | pass→pass | 17,071 | 16,592 | -3% | 1 | 1 | 0% | 2,835 | 4,126 | +46% | 0 | 0 | — |
case-10 | pass→fail | 8,001 | 3,550 | -56% | 1 | 1 | 0% | 1,407 | 1,747 | +24% | 0 | 0 | — |
case-11 | pass→pass | 13,057 | 16,000 | +23% | 1 | 1 | 0% | 2,302 | 4,000 | +74% | 0 | 0 | — |
case-12 | fail→pass | 13,327 | 11,173 | -16% | 1 | 1 | 0% | 2,419 | 3,304 | +37% | 0 | 0 | — |
case-13 | pass→pass | 10,557 | 12,999 | +23% | 1 | 1 | 0% | 2,121 | 3,539 | +67% | 0 | 0 | — |
case-14 | pass→pass | 16,512 | 18,179 | +10% | 1 | 1 | 0% | 2,831 | 4,398 | +55% | 0 | 0 | — |
case-15 | pass→pass | 15,766 | 20,277 | +29% | 1 | 1 | 0% | 2,740 | 4,680 | +71% | 0 | 0 | — |
case-16 | pass→pass | 12,681 | 11,896 | -6% | 1 | 1 | 0% | 2,358 | 3,350 | +42% | 0 | 0 | — |
case-17 | pass→pass | 6,686 | 8,810 | +32% | 1 | 1 | 0% | 1,156 | 2,656 | +130% | 0 | 0 | — |
case-18 | pass→pass | 17,761 | 17,963 | +1% | 1 | 1 | 0% | 3,202 | 4,630 | +45% | 0 | 0 | — |
case-19 | pass→pass | 17,908 | 19,658 | +10% | 1 | 1 | 0% | 3,014 | 4,566 | +51% | 0 | 0 | — |
case-20 | pass→pass | 15,002 | 24,562 | +64% | 1 | 1 | 0% | 2,973 | 5,582 | +88% | 0 | 0 | — |
case-21 | fail→pass | 18,587 | 18,844 | +1% | 1 | 1 | 0% | 3,150 | 4,542 | +44% | 0 | 0 | — |
case-22 | pass→pass | 17,000 | 14,444 | -15% | 1 | 1 | 0% | 2,603 | 3,728 | +43% | 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. The headline lift of +9 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.