Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Checks whether registered notebooks have current, stale, or missing outputs. Use before rendering or to verify freshness.
.claude/skills/brycewang-stanford-freeze-check/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -63% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -31% | 0% |
| case-18 | ✓→✓ | ＝ Same ✓ | -28% | 0% |
Verify that all registered notebooks have been executed and their outputs are current.
_quarto.yml and extract all notebook paths from manuscript.notebooks..ipynb JSON and check if any code cells have non-empty outputs arrays.ipynbexecution metadata timestamps in cell metadata (if present) to determine when outputs were last generated_freeze/notebooks/<name>/ directory exists and contains cached output.ipynb has not been modified since outputs were generated.ipynb source cells have been modified more recently than the outputs_freeze/ cache exists (Quarto will use the cache) Notebook Has Outputs Last Modified Status ───────────────────────────────────────────────────────────────── notebook-01.ipynb Yes 2026-02-28 14:30 Current notebook-02.ipynb Yes 2026-03-01 09:15 Stale notebook-03.ipynb No 2026-02-25 11:00 Unexecuted
/project:execute to re-execute all notebooks._quarto.yml has no notebooks registered, report "No notebooks found in _quarto.yml" and stop.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-18 | pass→pass | 7,365 | 2,097 | -72% | 1 | 1 | 0% | 1,228 | 883 | -28% | 0 | 0 | — |
case-01 | fail→fail | 16,617 | 2,654 | -84% | 1 | 1 | 0% | 3,395 | 765 | -77% | 0 | 0 | — |
case-02 | fail→fail | 2,619 | 6,810 | +160% | 1 | 1 | 0% | 387 | 849 | +119% | 0 | 0 | — |
case-03 | fail→fail | 14,578 | 2,306 | -84% | 1 | 1 | 0% | 2,912 | 771 | -74% | 0 | 0 | — |
case-04 | pass→pass | 9,385 | 2,991 | -68% | 1 | 1 | 0% | 1,572 | 968 | -38% | 0 | 0 | — |
case-05 | pass→pass | 8,189 | 3,221 | -61% | 1 | 1 | 0% | 1,471 | 1,085 | -26% | 0 | 0 | — |
case-06 | pass→pass | 8,660 | 2,509 | -71% | 1 | 1 | 0% | 1,579 | 917 | -42% | 0 | 0 | — |
case-07 | pass→pass | 7,443 | 2,069 | -72% | 1 | 1 | 0% | 1,189 | 835 | -30% | 0 | 0 | — |
case-08 | pass→fail | 7,620 | 1,930 | -75% | 1 | 1 | 0% | 1,120 | 776 | -31% | 0 | 0 | — |
case-09 | fail→fail | 10,523 | 2,810 | -73% | 1 | 1 | 0% | 1,910 | 839 | -56% | 0 | 0 | — |
case-10 | pass→pass | 10,008 | 4,101 | -59% | 1 | 1 | 0% | 1,837 | 1,203 | -35% | 0 | 0 | — |
case-11 | pass→pass | 3,725 | 1,651 | -56% | 1 | 1 | 0% | 678 | 725 | +7% | 0 | 0 | — |
case-12 | fail→pass | 10,034 | 5,642 | -44% | 1 | 1 | 0% | 1,841 | 1,423 | -23% | 0 | 0 | — |
case-13 | fail→pass | 12,022 | 1,844 | -85% | 1 | 1 | 0% | 2,057 | 764 | -63% | 0 | 0 | — |
case-14 | pass→pass | 5,204 | 5,809 | +12% | 1 | 1 | 0% | 1,056 | 1,475 | +40% | 0 | 0 | — |
case-15 | fail→pass | 5,496 | 2,270 | -59% | 1 | 1 | 0% | 951 | 768 | -19% | 0 | 0 | — |
case-16 | pass→pass | 6,529 | 2,680 | -59% | 1 | 1 | 0% | 1,130 | 824 | -27% | 0 | 0 | — |
case-17 | pass→pass | 5,512 | 3,585 | -35% | 1 | 1 | 0% | 1,115 | 1,091 | -2% | 0 | 0 | — |
case-19 | pass→pass | 8,956 | 2,321 | -74% | 1 | 1 | 0% | 1,497 | 884 | -41% | 0 | 0 | — |
case-20 | pass→pass | 7,849 | 16,135 | +106% | 1 | 1 | 0% | 1,438 | 2,648 | +84% | 0 | 0 | — |
case-21 | fail→fail | 3,229 | 4,816 | +49% | 1 | 1 | 0% | 495 | 694 | +40% | 0 | 0 | — |
case-22 | pass→pass | 6,135 | 4,371 | -29% | 1 | 1 | 0% | 1,141 | 1,260 | +10% | 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 +9 percentage points is the difference between those two pass rates over the 22 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.