Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Processes Excel spreadsheet files (.xlsx, .xlsm, .csv). Creates workbooks, builds formulas, preserves formatting, analyzes tabular data, and validates financial models with zero-formula-error delivery. Use when working with spreadsheet files or tabular data analysis. Do NOT use for Word documents, PDFs, presentations, or database pipelines.
.claude/skills/telagod-analyzing-spreadsheets/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-13 | ✓→✗ | ▼ Worse | -61% | 0% |
| case-17 | ✓→✗ | ▼ Worse | -41% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 95% | 0% |
| case-07 | ✓→✓ | ＝ Same ✓ | -33% | 0% |
Create, edit, analyze .xlsx files. LibreOffice required for formula recalculation via recalc.py.
Zero formula errors at delivery. All formulas must compute — no #REF!, #DIV/0!, #VALUE!, #N/A, #NAME?. Always run recalc.py after writing formulas.
| Task | Tool | Reference | |------|------|-----------| | Data analysis, bulk ops, simple export | pandas | recipes.md | | Formulas, formatting, Excel features | openpyxl | recipes.md | | Financial model standards | — | financial-model.md | | Recalculate formulas | recalc.py | recipes.md |
python recalc.py output.xlsx#REF! / #DIV/0! / #VALUE! / #NAME?| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 17,659 | 54,272 | +207% | 1 | 1 | 0% | 3,396 | 12,538 | +269% | 0 | 0 | — |
case-02 | fail→fail | 8,660 | 6,924 | -20% | 1 | 1 | 0% | 1,579 | 803 | -49% | 0 | 0 | — |
case-03 | fail→fail | 15,384 | 7,519 | -51% | 1 | 1 | 0% | 2,485 | 745 | -70% | 0 | 0 | — |
case-04 | fail→fail | 23,258 | 5,342 | -77% | 1 | 1 | 0% | 4,994 | 702 | -86% | 0 | 0 | — |
case-05 | pass→pass | 6,213 | 10,662 | +72% | 1 | 1 | 0% | 1,221 | 2,386 | +95% | 0 | 0 | — |
case-06 | fail→fail | 21,768 | 19,148 | -12% | 1 | 1 | 0% | 4,359 | 3,407 | -22% | 0 | 0 | — |
case-07 | pass→pass | 11,795 | 5,010 | -58% | 1 | 1 | 0% | 1,858 | 1,242 | -33% | 0 | 0 | — |
case-08 | pass→pass | 13,434 | 7,800 | -42% | 1 | 1 | 0% | 2,008 | 1,731 | -14% | 0 | 0 | — |
case-09 | fail→fail | 13,714 | 9,445 | -31% | 1 | 1 | 0% | 2,526 | 2,136 | -15% | 0 | 0 | — |
case-10 | pass→pass | 16,471 | 11,748 | -29% | 1 | 1 | 0% | 2,825 | 2,489 | -12% | 0 | 0 | — |
case-11 | pass→pass | 9,065 | 3,180 | -65% | 1 | 1 | 0% | 1,600 | 941 | -41% | 0 | 0 | — |
case-12 | fail→pass | 11,651 | 1,954 | -83% | 1 | 1 | 0% | 1,889 | 725 | -62% | 0 | 0 | — |
case-13 | pass→fail | 13,619 | 2,287 | -83% | 1 | 1 | 0% | 1,947 | 767 | -61% | 0 | 0 | — |
case-14 | pass→pass | 19,650 | 11,878 | -40% | 1 | 1 | 0% | 3,026 | 2,161 | -29% | 0 | 0 | — |
case-15 | pass→pass | 9,343 | 4,445 | -52% | 1 | 1 | 0% | 1,589 | 1,138 | -28% | 0 | 0 | — |
case-16 | pass→pass | 12,737 | 3,778 | -70% | 1 | 1 | 0% | 1,887 | 987 | -48% | 0 | 0 | — |
case-17 | pass→fail | 10,233 | 3,382 | -67% | 1 | 1 | 0% | 1,550 | 911 | -41% | 0 | 0 | — |
case-18 | pass→pass | 5,430 | 1,332 | -75% | 1 | 1 | 0% | 825 | 630 | -24% | 0 | 0 | — |
case-19 | pass→pass | 15,157 | 9,588 | -37% | 1 | 1 | 0% | 2,397 | 2,107 | -12% | 0 | 0 | — |
case-20 | pass→pass | 14,697 | 16,933 | +15% | 1 | 1 | 0% | 2,808 | 3,323 | +18% | 0 | 0 | — |
case-21 | pass→pass | 11,501 | 8,658 | -25% | 1 | 1 | 0% | 2,384 | 2,136 | -10% | 0 | 0 | — |
case-22 | pass→pass | 8,838 | 3,755 | -58% | 1 | 1 | 0% | 1,587 | 1,105 | -30% | 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 19 counted toward the lift figure. The other 3 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 -5 percentage points is the difference between those two pass rates over the 19 comparable cases. 4 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.