Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Wenn es um Aktenstruktur Und Dokumentenintake in Beamtenrecht geht: ordnet Akteninhalt, Belege, Lücken und Nachforderungen; liefert eine Schnittstellenkarte mit Kollisions-, Zuständigkeits- und Nachweisfragen.
.claude/skills/klotzkette-aktenstruktur-und-dokumentenintake/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -24% | 0% |
Beginne nicht mit einem Fragenkatalog. Wenn Material vorliegt, lies es zuerst und starte mit einer verwertbaren Arbeitshypothese:
Frage höchstens zwei Punkte nach, und nur wenn ohne diese Antwort der nächste Schritt falsch oder riskant würde. Fehlt Material vollständig, verlange nicht allgemein alle Unterlagen, sondern nenne die drei wichtigsten Dokumente und arbeite mit sichtbaren Annahmen weiter.
Starte mit einem Arbeitsprodukt, nicht mit einer Inventarliste: Kurzvermerk, Fristenblatt, Prüfmatrix, Entwurf, Fragenliste oder Entscheidungsvorschlag. Routing ist nur Mittel zum Zweck. Wenn ein Fachskill eindeutig passt, arbeite unmittelbar in dessen Richtung weiter.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 25,496 | 21,605 | -15% | 1 | 1 | 0% | 2,996 | 3,313 | +11% | 0 | 0 | — |
case-02 | fail→fail | 31,137 | 34,700 | +11% | 1 | 1 | 0% | 3,649 | 3,209 | -12% | 0 | 0 | — |
case-03 | fail→pass | 29,190 | 24,575 | -16% | 1 | 1 | 0% | 3,233 | 3,800 | +18% | 0 | 0 | — |
case-04 | fail→fail | 22,002 | 23,251 | +6% | 1 | 1 | 0% | 2,974 | 2,970 | -0% | 0 | 0 | — |
case-05 | pass→pass | 17,719 | 10,538 | -41% | 1 | 1 | 0% | 1,761 | 2,151 | +22% | 0 | 0 | — |
case-06 | pass→pass | 16,541 | 17,124 | +4% | 1 | 1 | 0% | 2,313 | 3,013 | +30% | 0 | 0 | — |
case-07 | pass→pass | 17,577 | 10,835 | -38% | 1 | 1 | 0% | 2,254 | 2,038 | -10% | 0 | 0 | — |
case-08 | pass→pass | 10,358 | 12,107 | +17% | 1 | 1 | 0% | 1,417 | 2,272 | +60% | 0 | 0 | — |
case-18 | fail→pass | 13,641 | 6,853 | -50% | 1 | 1 | 0% | 1,945 | 1,729 | -11% | 0 | 0 | — |
case-09 | fail→fail | 13,671 | 9,519 | -30% | 1 | 1 | 0% | 1,830 | 2,131 | +16% | 0 | 0 | — |
case-10 | pass→fail | 16,686 | 19,353 | +16% | 1 | 1 | 0% | 2,300 | 3,275 | +42% | 0 | 0 | — |
case-11 | pass→fail | 7,578 | 11,780 | +55% | 1 | 1 | 0% | 1,444 | 2,357 | +63% | 0 | 0 | — |
case-12 | pass→pass | 20,970 | 19,328 | -8% | 1 | 1 | 0% | 2,875 | 3,149 | +10% | 0 | 0 | — |
case-19 | pass→pass | 12,033 | 15,868 | +32% | 1 | 1 | 0% | 1,943 | 2,691 | +38% | 0 | 0 | — |
case-13 | fail→pass | 20,412 | 11,882 | -42% | 1 | 1 | 0% | 2,863 | 2,526 | -12% | 0 | 0 | — |
case-14 | fail→pass | 19,429 | 10,614 | -45% | 1 | 1 | 0% | 2,629 | 1,997 | -24% | 0 | 0 | — |
case-15 | pass→pass | 19,010 | 16,989 | -11% | 1 | 1 | 0% | 3,068 | 3,200 | +4% | 0 | 0 | — |
case-16 | fail→pass | 20,278 | 9,029 | -55% | 1 | 1 | 0% | 3,130 | 1,857 | -41% | 0 | 0 | — |
case-17 | pass→pass | 14,267 | 4,260 | -70% | 1 | 1 | 0% | 2,098 | 1,258 | -40% | 0 | 0 | — |
case-20 | fail→fail | 29,605 | 26,657 | -10% | 1 | 1 | 0% | 5,141 | 5,051 | -2% | 0 | 0 | — |
case-21 | fail→fail | 22,879 | 15,815 | -31% | 1 | 1 | 0% | 4,345 | 3,425 | -21% | 0 | 0 | — |
case-22 | pass→pass | 26,318 | 23,985 | -9% | 1 | 1 | 0% | 3,996 | 4,258 | +7% | 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 +18 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.