Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →You are the **Sales Data Extraction Agent** — an intelligent data pipeline specialist who monitors, parses, and extracts sales metrics from Excel files in real time. You are meticulous, accurate, a...
.claude/skills/dev-dennis-040-sales-data-extraction-agent/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -14% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 22% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 9% | 0% |
name: Sales Data Extraction Agent description: AI agent specialized in monitoring Excel files and extracting key sales metrics (MTD, YTD, Year End) for internal live reporting color: "#2b6cb0"
You are the Sales Data Extraction Agent — an intelligent data pipeline specialist who monitors, parses, and extracts sales metrics from Excel files in real time. You are meticulous, accurate, and never drop a data point.
Core Traits:
Monitor designated Excel file directories for new or updated sales reports. Extract key metrics — Month to Date (MTD), Year to Date (YTD), and Year End projections — then normalize and persist them for downstream reporting and distribution.
.xlsx and .xls files using filesystem watchers~$)revenue/sales/total_sales, units/qty/quantity, etc.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,556 | 9,933 | +52% | 1 | 1 | 0% | 1,517 | 2,951 | +95% | 0 | 0 | — |
case-02 | fail→fail | 13,467 | 13,952 | +4% | 1 | 1 | 0% | 3,197 | 3,931 | +23% | 0 | 0 | — |
case-03 | fail→pass | 13,989 | 12,741 | -9% | 1 | 1 | 0% | 3,650 | 3,777 | +3% | 0 | 0 | — |
case-04 | pass→pass | 14,151 | 12,743 | -10% | 1 | 1 | 0% | 2,985 | 3,637 | +22% | 0 | 0 | — |
case-05 | pass→pass | 22,651 | 28,330 | +25% | 1 | 1 | 0% | 6,172 | 6,745 | +9% | 0 | 0 | — |
case-06 | pass→fail | 10,291 | 6,885 | -33% | 1 | 1 | 0% | 2,130 | 1,826 | -14% | 0 | 0 | — |
case-07 | pass→pass | 10,351 | 8,998 | -13% | 1 | 1 | 0% | 1,963 | 2,407 | +23% | 0 | 0 | — |
case-08 | pass→pass | 8,917 | 6,412 | -28% | 1 | 1 | 0% | 1,719 | 1,962 | +14% | 0 | 0 | — |
case-09 | pass→pass | 8,458 | 6,285 | -26% | 1 | 1 | 0% | 1,738 | 1,711 | -2% | 0 | 0 | — |
case-10 | fail→pass | 12,301 | 9,802 | -20% | 1 | 1 | 0% | 2,770 | 2,684 | -3% | 0 | 0 | — |
case-11 | pass→pass | 11,894 | 10,610 | -11% | 1 | 1 | 0% | 2,451 | 2,840 | +16% | 0 | 0 | — |
case-12 | pass→pass | 10,895 | 9,680 | -11% | 1 | 1 | 0% | 2,095 | 2,547 | +22% | 0 | 0 | — |
case-13 | pass→pass | 12,601 | 12,019 | -5% | 1 | 1 | 0% | 2,819 | 3,142 | +11% | 0 | 0 | — |
case-14 | fail→fail | 9,288 | 8,567 | -8% | 1 | 1 | 0% | 1,875 | 2,158 | +15% | 0 | 0 | — |
case-15 | pass→pass | 7,557 | 4,882 | -35% | 1 | 1 | 0% | 1,656 | 1,576 | -5% | 0 | 0 | — |
case-16 | pass→pass | 10,171 | 4,755 | -53% | 1 | 1 | 0% | 1,909 | 1,414 | -26% | 0 | 0 | — |
case-17 | pass→pass | 7,558 | 3,643 | -52% | 1 | 1 | 0% | 1,465 | 1,222 | -17% | 0 | 0 | — |
case-18 | pass→pass | 12,201 | 7,528 | -38% | 1 | 1 | 0% | 2,294 | 2,032 | -11% | 0 | 0 | — |
case-19 | pass→pass | 9,086 | 3,865 | -57% | 1 | 1 | 0% | 1,859 | 1,215 | -35% | 0 | 0 | — |
case-20 | pass→pass | 10,737 | 7,688 | -28% | 1 | 1 | 0% | 2,046 | 2,188 | +7% | 0 | 0 | — |
case-21 | pass→pass | 10,910 | 8,524 | -22% | 1 | 1 | 0% | 2,033 | 2,259 | +11% | 0 | 0 | — |
case-22 | pass→pass | 10,212 | 7,433 | -27% | 1 | 1 | 0% | 1,932 | 2,069 | +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 +5 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.