Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Scaffolds a method-specific analysis notebook (DiD, IV, RDD, LASSO, Panel FE) with boilerplate. Use when starting a new econometric analysis.
.claude/skills/brycewang-stanford-new-analysis/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-15 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -17% | 0% |
Create a new notebook pre-populated with method-specific boilerplate for a common econometric technique.
$ARGUMENTS — the method name and optional title (e.g., "DiD Event Study", "IV Analysis of Colonial Origins", "RDD Minimum Wage", "LASSO Variable Selection", "Panel FE Growth Regressions")/project:new-notebook:notebooks/ for existing files to determine the next sequential number.ipynb with the appropriate kernel and setup cell:import sys; sys.path.insert(0, ".."); from config import set_seeds, DATA_DIR; set_seeds()source("../config.R"); set_seeds()clear all followed by set seed 42All methods include these sections:
#| label: tbl-<method>-sumstats)#| label: tbl-<method>-main)#| label: fig-<method>-main)Method-specific boilerplate:
#| label: fig-event-study), TWFE regression, staggered treatment note#| label: fig-rd-plot)#| label: fig-lasso-path), selected variables, post-LASSO OLS.md pair:bash uv run jupytext --set-formats ipynb,md:myst notebooks/<name>.ipynb
_quarto.yml under manuscript.notebooks:yaml
title: "N<number>: <title>"
quarto render notebooks/<name>.ipynb| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | fail→pass | 18,283 | 9,944 | -46% | 1 | 1 | 0% | 2,874 | 2,452 | -15% | 0 | 0 | — |
case-03 | fail→fail | 27,283 | 6,292 | -77% | 1 | 1 | 0% | 5,390 | 1,037 | -81% | 0 | 0 | — |
case-01 | fail→fail | 35,639 | 5,708 | -84% | 1 | 1 | 0% | 4,816 | 943 | -80% | 0 | 0 | — |
case-02 | fail→fail | 40,956 | 5,531 | -86% | 1 | 1 | 0% | 4,676 | 973 | -79% | 0 | 0 | — |
case-09 | fail→fail | 20,336 | 5,638 | -72% | 1 | 1 | 0% | 3,820 | 996 | -74% | 0 | 0 | — |
case-04 | fail→fail | 14,344 | 5,189 | -64% | 1 | 1 | 0% | 3,037 | 1,004 | -67% | 0 | 0 | — |
case-05 | fail→fail | 25,012 | 5,652 | -77% | 1 | 1 | 0% | 5,300 | 916 | -83% | 0 | 0 | — |
case-06 | fail→pass | 22,115 | 2,959 | -87% | 1 | 1 | 0% | 1,098 | 1,263 | +15% | 0 | 0 | — |
case-07 | fail→pass | 11,277 | 7,714 | -32% | 1 | 1 | 0% | 1,947 | 1,693 | -13% | 0 | 0 | — |
case-08 | fail→pass | 13,493 | 6,164 | -54% | 1 | 1 | 0% | 2,128 | 1,619 | -24% | 0 | 0 | — |
case-10 | fail→pass | 10,489 | 4,902 | -53% | 1 | 1 | 0% | 1,858 | 1,542 | -17% | 0 | 0 | — |
case-11 | pass→pass | 3,652 | 3,254 | -11% | 1 | 1 | 0% | 566 | 1,306 | +131% | 0 | 0 | — |
case-12 | fail→pass | 4,628 | 2,944 | -36% | 1 | 1 | 0% | 844 | 1,219 | +44% | 0 | 0 | — |
case-13 | pass→pass | 15,698 | 8,419 | -46% | 1 | 1 | 0% | 2,667 | 2,002 | -25% | 0 | 0 | — |
case-14 | pass→pass | 18,532 | 4,756 | -74% | 1 | 1 | 0% | 3,279 | 1,480 | -55% | 0 | 0 | — |
case-16 | fail→pass | 11,140 | 2,675 | -76% | 1 | 1 | 0% | 1,887 | 1,178 | -38% | 0 | 0 | — |
case-17 | pass→pass | 4,551 | 2,716 | -40% | 1 | 1 | 0% | 479 | 1,006 | +110% | 0 | 0 | — |
case-18 | fail→pass | 12,702 | 7,528 | -41% | 1 | 1 | 0% | 1,981 | 2,082 | +5% | 0 | 0 | — |
case-19 | pass→pass | 14,858 | 17,743 | +19% | 1 | 1 | 0% | 2,749 | 3,492 | +27% | 0 | 0 | — |
case-20 | pass→pass | 12,664 | 12,012 | -5% | 1 | 1 | 0% | 2,246 | 2,864 | +28% | 0 | 0 | — |
case-21 | pass→pass | 14,503 | 11,783 | -19% | 1 | 1 | 0% | 2,484 | 2,836 | +14% | 0 | 0 | — |
case-22 | fail→pass | 13,924 | 3,774 | -73% | 1 | 1 | 0% | 2,252 | 1,403 | -38% | 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 15 counted toward the lift figure. The other 7 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 +41 percentage points is the difference between those two pass rates over the 15 comparable cases. 3 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.