Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create publication-quality charts and graphs for economics papers.
.claude/skills/brycewang-stanford-econ-visualization/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 71% | 0% |
| case-19 | ✓→✓ | ＝ Same ✓ | 6% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 44% | 0% |
This skill creates publication-quality figures for economics papers, using clean styling, consistent scales, and export-ready formats.
Follow these steps to complete the task:
Before generating any code, ask the user:
Based on the context, generate code that:
After generating output:
r# ============================================ # Publication-Quality Figure in R # ============================================ library(tidyverse) df <- read_csv("data.csv") ggplot(df, aes(x = year, y = gdp_per_capita, color = country)) + geom_line(size = 1) + scale_y_continuous(labels = scales::comma) + labs( title = "GDP per Capita Over Time", x = "Year", y = "GDP per Capita (USD)", color = "Country" ) + theme_minimal(base_size = 12) + theme( legend.position = "bottom", panel.grid.minor = element_blank() ) ggsave("figures/gdp_per_capita.pdf", width = 7, height = 4, dpi = 300)
ggplot2, scales, dplyrmatplotlib, seaborn (optional alternative)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | pass→pass | 14,969 | 13,277 | -11% | 1 | 1 | 0% | 3,358 | 3,543 | +6% | 0 | 0 | — |
case-01 | fail→fail | 16,533 | 15,124 | -9% | 1 | 1 | 0% | 3,204 | 3,672 | +15% | 0 | 0 | — |
case-02 | fail→fail | 23,036 | 17,730 | -23% | 1 | 1 | 0% | 4,323 | 4,263 | -1% | 0 | 0 | — |
case-03 | fail→fail | 18,556 | 17,385 | -6% | 1 | 1 | 0% | 3,831 | 4,048 | +6% | 0 | 0 | — |
case-04 | pass→pass | 9,146 | 10,068 | +10% | 1 | 1 | 0% | 1,872 | 2,688 | +44% | 0 | 0 | — |
case-05 | fail→pass | 9,527 | 10,734 | +13% | 1 | 1 | 0% | 2,081 | 3,034 | +46% | 0 | 0 | — |
case-06 | pass→pass | 13,404 | 11,013 | -18% | 1 | 1 | 0% | 2,770 | 2,868 | +4% | 0 | 0 | — |
case-07 | fail→fail | 19,775 | 11,490 | -42% | 1 | 1 | 0% | 4,208 | 2,891 | -31% | 0 | 0 | — |
case-08 | pass→pass | 12,310 | 12,238 | -1% | 1 | 1 | 0% | 2,524 | 2,974 | +18% | 0 | 0 | — |
case-09 | pass→pass | 14,325 | 12,384 | -14% | 1 | 1 | 0% | 2,650 | 3,043 | +15% | 0 | 0 | — |
case-10 | pass→pass | 5,605 | 10,691 | +91% | 1 | 1 | 0% | 830 | 2,905 | +250% | 0 | 0 | — |
case-11 | pass→pass | 12,353 | 14,324 | +16% | 1 | 1 | 0% | 2,450 | 3,770 | +54% | 0 | 0 | — |
case-12 | fail→fail | 20,255 | 14,418 | -29% | 1 | 1 | 0% | 3,899 | 3,548 | -9% | 0 | 0 | — |
case-13 | fail→pass | 15,879 | 12,046 | -24% | 1 | 1 | 0% | 3,151 | 3,183 | +1% | 0 | 0 | — |
case-14 | fail→pass | 6,736 | 7,566 | +12% | 1 | 1 | 0% | 1,306 | 2,238 | +71% | 0 | 0 | — |
case-15 | pass→pass | 7,810 | 8,137 | +4% | 1 | 1 | 0% | 1,538 | 2,324 | +51% | 0 | 0 | — |
case-16 | pass→pass | 10,243 | 9,831 | -4% | 1 | 1 | 0% | 1,843 | 2,601 | +41% | 0 | 0 | — |
case-17 | pass→pass | 12,893 | 12,611 | -2% | 1 | 1 | 0% | 2,507 | 2,960 | +18% | 0 | 0 | — |
case-18 | fail→fail | 18,207 | 14,140 | -22% | 1 | 1 | 0% | 3,912 | 3,688 | -6% | 0 | 0 | — |
case-20 | fail→fail | 16,965 | 13,962 | -18% | 1 | 1 | 0% | 3,176 | 3,503 | +10% | 0 | 0 | — |
case-21 | fail→fail | 21,375 | 18,570 | -13% | 1 | 1 | 0% | 3,993 | 4,344 | +9% | 0 | 0 | — |
case-22 | pass→pass | 8,527 | 10,262 | +20% | 1 | 1 | 0% | 1,582 | 2,851 | +80% | 0 | 0 | — |
case-23 | pass→pass | 11,473 | 11,664 | +2% | 1 | 1 | 0% | 2,246 | 3,025 | +35% | 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. The headline lift of +13 percentage points is the difference between those two pass rates over the 23 comparable cases.
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.