Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Read a CSV file and produce a concise summary including row count, column names, inferred types per column, and any obvious data-quality issues. Activates when the user asks to summarize, profile, or describe a CSV file or any tabular dataset in the working directory.
.claude/skills/suifeng156-csv-summarizer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 57% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 113% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 155% | 0% |
When the user asks you to summarize or profile a CSV file, follow this protocol.
If the user names a specific file, use that. Otherwise list .csv files in the working directory and use the most recently modified one. Report the filename you chose.
Read the file (or its first 1000 rows if very large). Compute:
integer, decimal, date, boolean, string, or mixed.Produce a single response with:
csv-summarizer: <filename> (N rows × M columns).| name | inferred type | sample | notes |.Do not modify the file. Read-only.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | fail→fail | 7,562 | 9,051 | +20% | 1 | 1 | 0% | 1,494 | 868 | -42% | 0 | 0 | — |
case-01 | fail→fail | 17,539 | 6,374 | -64% | 1 | 1 | 0% | 3,381 | 767 | -77% | 0 | 0 | — |
case-02 | fail→fail | 9,134 | 7,022 | -23% | 1 | 1 | 0% | 550 | 583 | +6% | 0 | 0 | — |
case-03 | fail→fail | 13,991 | 7,564 | -46% | 1 | 1 | 0% | 2,730 | 752 | -72% | 0 | 0 | — |
case-04 | fail→fail | 12,418 | 6,562 | -47% | 1 | 1 | 0% | 2,397 | 720 | -70% | 0 | 0 | — |
case-05 | fail→pass | 12,073 | 20,232 | +68% | 1 | 1 | 0% | 2,192 | 3,441 | +57% | 0 | 0 | — |
case-06 | fail→pass | 8,381 | 21,402 | +155% | 1 | 1 | 0% | 1,520 | 3,237 | +113% | 0 | 0 | — |
case-08 | fail→fail | 11,634 | 9,548 | -18% | 1 | 1 | 0% | 2,366 | 907 | -62% | 0 | 0 | — |
case-09 | fail→pass | 16,566 | 16,297 | -2% | 1 | 1 | 0% | 3,193 | 2,551 | -20% | 0 | 0 | — |
case-10 | fail→pass | 13,267 | 20,191 | +52% | 1 | 1 | 0% | 2,390 | 3,793 | +59% | 0 | 0 | — |
case-11 | pass→pass | 18,700 | 23,322 | +25% | 1 | 1 | 0% | 3,546 | 3,670 | +3% | 0 | 0 | — |
case-17 | fail→fail | 7,206 | 9,004 | +25% | 1 | 1 | 0% | 1,379 | 662 | -52% | 0 | 0 | — |
case-12 | fail→pass | 5,438 | 10,681 | +96% | 1 | 1 | 0% | 937 | 2,389 | +155% | 0 | 0 | — |
case-13 | fail→fail | 6,538 | 7,374 | +13% | 1 | 1 | 0% | 1,221 | 663 | -46% | 0 | 0 | — |
case-14 | fail→fail | 1,931 | 7,528 | +290% | 1 | 1 | 0% | 283 | 589 | +108% | 0 | 0 | — |
case-15 | fail→pass | 10,859 | 15,767 | +45% | 1 | 1 | 0% | 1,572 | 3,295 | +110% | 0 | 0 | — |
case-16 | pass→fail | 11,065 | 5,580 | -50% | 1 | 1 | 0% | 2,187 | 670 | -69% | 0 | 0 | — |
case-18 | fail→fail | 16,907 | 10,890 | -36% | 1 | 1 | 0% | 1,963 | 791 | -60% | 0 | 0 | — |
case-19 | fail→fail | 16,506 | 7,301 | -56% | 1 | 1 | 0% | 3,166 | 808 | -74% | 0 | 0 | — |
case-20 | pass→pass | 3,985 | 2,885 | -28% | 1 | 1 | 0% | 640 | 807 | +26% | 0 | 0 | — |
case-21 | pass→pass | 4,311 | 2,448 | -43% | 1 | 1 | 0% | 764 | 716 | -6% | 0 | 0 | — |
case-22 | pass→pass | 3,715 | 4,303 | +16% | 1 | 1 | 0% | 723 | 1,082 | +50% | 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 10 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 +23 percentage points is the difference between those two pass rates over the 10 comparable cases. 2 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.