Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Wenn es um Arbeitsrecht - Problem Sortieren in Arbeitsrecht geht: prüft Frist, Form, Zuständigkeit, Rechtsweg und Sofortmaßnahmen; liefert eine Fristen- und Risikoampel mit Sofortschritten.
.claude/skills/klotzkette-arbeitsrecht-problem-sortieren/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 50% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 42% | 0% |
Dieser Skill ist der ruhige erste Tisch in einer unuebersichtlichen arbeitsrechtlichen Lage. Er beginnt nicht mit einem Schema, sondern mit Ordnung: Was ist passiert, was will die Person erreichen, was liegt als Dokument vor, was eilt und welcher Spezial-Skill hilft jetzt wirklich?
Schnellfeuer-Klage spart Frist, aber ohne Sozialauswahl-Prüfung und ohne Betriebsratsanhörungs-Check verschenkt Mandantschaft Argumente. Lieber 3-Wochen-Frist sichern mit knapper Klageschrift, dann nachschriftsätzlich substantiieren.
[noch klaeren] markieren.| Punkt | Befund | Risiko | Naechster Schritt | | --- | --- | --- | --- | | ... | ... | ... | ... |
Ist die Antwort handlungsorientiert, knapp, respektvoll, belegnah und ohne erfundene Quellen? Sind Fristen und offene Tatsachen sichtbar? Ist der naechste Schritt eindeutig?
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | pass→pass | 12,357 | 14,943 | +21% | 1 | 1 | 0% | 2,376 | 3,981 | +68% | 0 | 0 | — |
case-01 | fail→fail | 18,478 | 14,216 | -23% | 1 | 1 | 0% | 3,398 | 4,185 | +23% | 0 | 0 | — |
case-02 | fail→fail | 17,035 | 19,657 | +15% | 1 | 1 | 0% | 3,321 | 4,914 | +48% | 0 | 0 | — |
case-03 | fail→pass | 15,893 | 19,730 | +24% | 1 | 1 | 0% | 3,041 | 4,744 | +56% | 0 | 0 | — |
case-04 | pass→pass | 10,388 | 13,553 | +30% | 1 | 1 | 0% | 2,402 | 3,755 | +56% | 0 | 0 | — |
case-05 | pass→pass | 8,169 | 13,648 | +67% | 1 | 1 | 0% | 1,708 | 4,036 | +136% | 0 | 0 | — |
case-06 | pass→pass | 6,311 | 12,449 | +97% | 1 | 1 | 0% | 1,312 | 3,420 | +161% | 0 | 0 | — |
case-07 | pass→pass | 8,880 | 12,038 | +36% | 1 | 1 | 0% | 1,589 | 3,590 | +126% | 0 | 0 | — |
case-08 | pass→pass | 7,705 | 13,462 | +75% | 1 | 1 | 0% | 1,441 | 3,614 | +151% | 0 | 0 | — |
case-09 | pass→pass | 10,197 | 13,358 | +31% | 1 | 1 | 0% | 1,685 | 3,636 | +116% | 0 | 0 | — |
case-10 | pass→pass | 8,020 | 8,418 | +5% | 1 | 1 | 0% | 1,456 | 2,634 | +81% | 0 | 0 | — |
case-11 | pass→pass | 10,054 | 15,940 | +59% | 1 | 1 | 0% | 1,888 | 4,054 | +115% | 0 | 0 | — |
case-12 | pass→pass | 11,509 | 14,243 | +24% | 1 | 1 | 0% | 1,967 | 3,715 | +89% | 0 | 0 | — |
case-13 | pass→pass | 10,008 | 16,077 | +61% | 1 | 1 | 0% | 1,676 | 3,915 | +134% | 0 | 0 | — |
case-14 | pass→pass | 12,757 | 17,411 | +36% | 1 | 1 | 0% | 2,260 | 3,923 | +74% | 0 | 0 | — |
case-15 | fail→pass | 17,292 | 13,692 | -21% | 1 | 1 | 0% | 2,901 | 3,969 | +37% | 0 | 0 | — |
case-16 | fail→pass | 13,982 | 13,036 | -7% | 1 | 1 | 0% | 2,265 | 3,390 | +50% | 0 | 0 | — |
case-18 | fail→pass | 9,342 | 10,397 | +11% | 1 | 1 | 0% | 2,205 | 3,079 | +40% | 0 | 0 | — |
case-19 | fail→pass | 9,911 | 8,891 | -10% | 1 | 1 | 0% | 1,966 | 2,786 | +42% | 0 | 0 | — |
case-20 | fail→fail | 19,575 | 17,087 | -13% | 1 | 1 | 0% | 3,538 | 4,485 | +27% | 0 | 0 | — |
case-21 | fail→pass | 18,521 | 15,200 | -18% | 1 | 1 | 0% | 3,562 | 3,913 | +10% | 0 | 0 | — |
case-22 | fail→fail | 16,278 | 17,064 | +5% | 1 | 1 | 0% | 3,277 | 4,400 | +34% | 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 +27 percentage points is the difference between those two pass rates over the 22 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.