Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →6 mathematics skills. Trigger: mathematical proofs, theorem proving, numerical methods, linear algebra. Design: formal verification tools and computational mathematics guides.
.claude/skills/brycewang-stanford-math-skills/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -74% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -82% | 0% |
Select the skill matching the user's need, then read its SKILL.md.
| Skill | Description | |-------|-------------| | lean-theorem-proving-guide | LLM agent for formal theorem proving in Lean 4 | | linear-algebra-applications | Apply linear algebra concepts to research computing and data analysis | | numerical-methods-guide | Apply numerical methods and scientific computing techniques | | oeis-api | On-Line Encyclopedia of Integer Sequences API | | symbolic-computation-guide | Computer algebra systems: SymPy, SageMath, and Mathematica for research | | topology-data-analysis | Topological data analysis: persistent homology, Mapper, and TDA tools |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,251 | 5,774 | -74% | 1 | 1 | 0% | 3,612 | 391 | -89% | 0 | 0 | — |
case-02 | fail→fail | 17,224 | 4,456 | -74% | 1 | 1 | 0% | 2,342 | 452 | -81% | 0 | 0 | — |
case-03 | fail→pass | 6,923 | 1,977 | -71% | 1 | 1 | 0% | 996 | 591 | -41% | 0 | 0 | — |
case-04 | fail→pass | 14,791 | 2,542 | -83% | 1 | 1 | 0% | 2,419 | 625 | -74% | 0 | 0 | — |
case-05 | fail→pass | 7,043 | 6,881 | -2% | 1 | 1 | 0% | 1,032 | 690 | -33% | 0 | 0 | — |
case-06 | fail→pass | 12,569 | 13,164 | +5% | 1 | 1 | 0% | 2,206 | 1,768 | -20% | 0 | 0 | — |
case-07 | fail→pass | 16,097 | 5,572 | -65% | 1 | 1 | 0% | 2,338 | 420 | -82% | 0 | 0 | — |
case-08 | fail→fail | 35,085 | 5,110 | -85% | 1 | 1 | 0% | 2,734 | 586 | -79% | 0 | 0 | — |
case-09 | fail→fail | 6,625 | 5,943 | -10% | 1 | 1 | 0% | 819 | 665 | -19% | 0 | 0 | — |
case-10 | fail→fail | 12,735 | 15,196 | +19% | 1 | 1 | 0% | 2,185 | 637 | -71% | 0 | 0 | — |
case-11 | fail→fail | 6,068 | 3,945 | -35% | 1 | 1 | 0% | 845 | 860 | +2% | 0 | 0 | — |
case-12 | pass→fail | 6,045 | 3,900 | -35% | 1 | 1 | 0% | 973 | 544 | -44% | 0 | 0 | — |
case-13 | pass→fail | 5,600 | 2,702 | -52% | 1 | 1 | 0% | 1,016 | 400 | -61% | 0 | 0 | — |
case-14 | pass→fail | 8,987 | 34,434 | +283% | 1 | 1 | 0% | 1,596 | 520 | -67% | 0 | 0 | — |
case-15 | pass→fail | 7,415 | 6,165 | -17% | 1 | 1 | 0% | 1,101 | 698 | -37% | 0 | 0 | — |
case-16 | pass→pass | 5,066 | 7,955 | +57% | 1 | 1 | 0% | 834 | 1,514 | +82% | 0 | 0 | — |
case-17 | pass→fail | 13,118 | 2,647 | -80% | 1 | 1 | 0% | 2,138 | 404 | -81% | 0 | 0 | — |
case-18 | pass→pass | 4,849 | 4,189 | -14% | 1 | 1 | 0% | 723 | 842 | +16% | 0 | 0 | — |
case-19 | pass→fail | 3,396 | 3,174 | -7% | 1 | 1 | 0% | 578 | 412 | -29% | 0 | 0 | — |
case-20 | pass→fail | 6,348 | 27,807 | +338% | 1 | 1 | 0% | 1,044 | 499 | -52% | 0 | 0 | — |
case-21 | pass→fail | 11,057 | 3,888 | -65% | 1 | 1 | 0% | 1,793 | 476 | -73% | 0 | 0 | — |
case-22 | pass→pass | 6,683 | 7,085 | +6% | 1 | 1 | 0% | 943 | 830 | -12% | 0 | 0 | — |
case-23 | pass→pass | 5,874 | 4,966 | -15% | 1 | 1 | 0% | 950 | 925 | -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. 23 cases were attempted, and 11 counted toward the lift figure. The other 12 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 -13 percentage points is the difference between those two pass rates over the 11 comparable cases. 11 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.