Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Captures tool versions, packages, and kernel info as a reproducibility record in notes/. Use when documenting the environment.
.claude/skills/brycewang-stanford-env-snapshot/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 103% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 200% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -60% | 0% |
Capture the full environment state and save it as a reproducibility record.
bash uname -a # OS info quarto --version # Quarto uv --version # uv python3 --version # Python
bash uv pip list # all installed packages with versions
bash uv run jupyter kernelspec list
bash R -e "sessionInfo()"
bash which stata # or check nbstata.conf for edition info Read ~/.config/nbstata/nbstata.conf for stata_dir and edition.
bash pdflatex --version | head -1
markdown # Environment Snapshot — YYYY-MM-DD
## System
## Tools | Tool | Version | | ---- | ------- | | Quarto | x.x.x | | uv | x.x.x | | Python | 3.12.x | | R | x.x.x (or N/A) | | Stata | SE x.x (or N/A) | | TeX Live | xxxx (or N/A) |
## Jupyter Kernels
## Python Packages | Package | Version | | ------- | ------- | | numpy | x.x.x | | pandas | x.x.x | | ... | ... |
## R Session Info (full sessionInfo() output, or "R not available")
## Stata Packages (ado dir output, or "Stata not available")
notes/environment-YYYYMMDD.md (using today's date).| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→pass | 5,631 | 1,890 | -66% | 1 | 1 | 0% | 1,006 | 913 | -9% | 0 | 0 | — |
case-01 | fail→fail | 7,458 | 8,230 | +10% | 1 | 1 | 0% | 1,516 | 979 | -35% | 0 | 0 | — |
case-02 | fail→fail | 13,868 | 5,203 | -62% | 1 | 1 | 0% | 3,175 | 912 | -71% | 0 | 0 | — |
case-03 | fail→fail | 8,303 | 5,596 | -33% | 1 | 1 | 0% | 1,737 | 968 | -44% | 0 | 0 | — |
case-04 | fail→pass | 4,464 | 9,811 | +120% | 1 | 1 | 0% | 880 | 1,787 | +103% | 0 | 0 | — |
case-05 | pass→fail | 6,276 | 7,116 | +13% | 1 | 1 | 0% | 1,127 | 1,000 | -11% | 0 | 0 | — |
case-06 | pass→pass | 6,236 | 6,455 | +4% | 1 | 1 | 0% | 1,388 | 1,769 | +27% | 0 | 0 | — |
case-07 | fail→pass | 8,750 | 5,546 | -37% | 1 | 1 | 0% | 1,472 | 1,446 | -2% | 0 | 0 | — |
case-08 | fail→pass | 2,263 | 2,905 | +28% | 1 | 1 | 0% | 331 | 993 | +200% | 0 | 0 | — |
case-10 | pass→pass | 5,143 | 2,131 | -59% | 1 | 1 | 0% | 976 | 897 | -8% | 0 | 0 | — |
case-11 | pass→pass | 9,016 | 6,269 | -30% | 1 | 1 | 0% | 1,687 | 1,673 | -1% | 0 | 0 | — |
case-12 | fail→pass | 12,147 | 1,887 | -84% | 1 | 1 | 0% | 2,130 | 843 | -60% | 0 | 0 | — |
case-13 | pass→pass | 6,711 | 2,037 | -70% | 1 | 1 | 0% | 1,257 | 878 | -30% | 0 | 0 | — |
case-14 | pass→pass | 3,051 | 4,003 | +31% | 1 | 1 | 0% | 525 | 1,302 | +148% | 0 | 0 | — |
case-15 | fail→pass | 10,833 | 4,447 | -59% | 1 | 1 | 0% | 2,131 | 1,391 | -35% | 0 | 0 | — |
case-16 | fail→pass | 5,088 | 2,402 | -53% | 1 | 1 | 0% | 674 | 1,052 | +56% | 0 | 0 | — |
case-17 | pass→pass | 9,242 | 3,693 | -60% | 1 | 1 | 0% | 1,810 | 1,287 | -29% | 0 | 0 | — |
case-18 | fail→pass | 5,918 | 1,538 | -74% | 1 | 1 | 0% | 974 | 789 | -19% | 0 | 0 | — |
case-19 | fail→pass | 5,673 | 1,710 | -70% | 1 | 1 | 0% | 956 | 796 | -17% | 0 | 0 | — |
case-20 | pass→pass | 2,093 | 1,226 | -41% | 1 | 1 | 0% | 348 | 739 | +112% | 0 | 0 | — |
case-21 | fail→pass | 4,755 | 2,986 | -37% | 1 | 1 | 0% | 801 | 1,088 | +36% | 0 | 0 | — |
case-22 | pass→fail | 5,651 | 1,683 | -70% | 1 | 1 | 0% | 850 | 882 | +4% | 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 18 counted toward the lift figure. The other 4 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 +36 percentage points is the difference between those two pass rates over the 18 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.