Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Jupyter ノートブック (.ipynb) を marimo ノートブック (.py) に変換するスキル。 「Jupyterを変換して」「ipynbをmarimoに」「ノートブック変換」等のリクエストで発動。
.claude/skills/minicoohei-jupyter-to-marimo/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 282% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -68% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -58% | 0% |
「Jupyter変換」「ipynb変換」「marimoに変換」「ノートブック変換」
IMPORTANT: When asked to translate a notebook, ALWAYS run uvx marimo convert <notebook.ipynb> -o <notebook.py> FIRST before reading any files. This saves precious tokens - reading large notebooks can consume 30k+ tokens, while the converted .py file is much smaller and easier to work with.
Run the marimo convert command via uvx so no install is needed:
bashuvx marimo convert <notebook.ipynb> -o <notebook.py>
This generates a marimo-compatible .py file from the Jupyter notebook.
marimo check on the outputbashuvx marimo check <notebook.py>
Fix any issues that are reported before continuing.
Read the generated .py file and apply the following improvements:
display() calls, or %magic commands that don't apply in marimo.EnvConfig widget from wigglystuff. Details can be found here.mo.ui.slider().marimo check again after your edits to confirm nothing was broken.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 23,102 | 3,525 | -85% | 1 | 1 | 0% | 5,050 | 1,027 | -80% | 0 | 0 | — |
case-02 | fail→fail | 23,415 | 5,243 | -78% | 1 | 1 | 0% | 4,960 | 650 | -87% | 0 | 0 | — |
case-03 | fail→pass | 6,281 | 16,855 | +168% | 1 | 1 | 0% | 1,058 | 4,044 | +282% | 0 | 0 | — |
case-04 | pass→pass | 8,651 | 5,414 | -37% | 1 | 1 | 0% | 1,664 | 1,423 | -14% | 0 | 0 | — |
case-05 | pass→pass | 4,549 | 1,944 | -57% | 1 | 1 | 0% | 807 | 706 | -13% | 0 | 0 | — |
case-06 | fail→pass | 11,905 | 1,844 | -85% | 1 | 1 | 0% | 2,160 | 684 | -68% | 0 | 0 | — |
case-07 | pass→pass | 6,773 | 4,081 | -40% | 1 | 1 | 0% | 1,238 | 1,209 | -2% | 0 | 0 | — |
case-08 | pass→pass | 10,137 | 3,310 | -67% | 1 | 1 | 0% | 1,785 | 975 | -45% | 0 | 0 | — |
case-09 | pass→pass | 9,819 | 6,364 | -35% | 1 | 1 | 0% | 1,757 | 1,568 | -11% | 0 | 0 | — |
case-10 | pass→pass | 9,931 | 6,908 | -30% | 1 | 1 | 0% | 1,565 | 1,610 | +3% | 0 | 0 | — |
case-11 | fail→pass | 10,306 | 8,485 | -18% | 1 | 1 | 0% | 1,705 | 1,948 | +14% | 0 | 0 | — |
case-12 | pass→pass | 8,252 | 2,849 | -65% | 1 | 1 | 0% | 1,507 | 914 | -39% | 0 | 0 | — |
case-13 | pass→pass | 14,581 | 2,540 | -83% | 1 | 1 | 0% | 1,109 | 780 | -30% | 0 | 0 | — |
case-14 | fail→pass | 12,318 | 4,664 | -62% | 1 | 1 | 0% | 1,950 | 1,241 | -36% | 0 | 0 | — |
case-15 | pass→pass | 3,427 | 3,974 | +16% | 1 | 1 | 0% | 589 | 1,148 | +95% | 0 | 0 | — |
case-16 | pass→pass | 9,727 | 6,562 | -33% | 1 | 1 | 0% | 1,669 | 1,422 | -15% | 0 | 0 | — |
case-17 | fail→pass | 9,331 | 1,975 | -79% | 1 | 1 | 0% | 1,674 | 709 | -58% | 0 | 0 | — |
case-22 | pass→pass | 3,591 | 3,256 | -9% | 1 | 1 | 0% | 568 | 911 | +60% | 0 | 0 | — |
case-18 | pass→pass | 14,334 | 6,776 | -53% | 1 | 1 | 0% | 2,427 | 1,553 | -36% | 0 | 0 | — |
case-19 | pass→pass | 8,780 | 5,430 | -38% | 1 | 1 | 0% | 1,340 | 1,241 | -7% | 0 | 0 | — |
case-20 | pass→pass | 6,022 | 4,248 | -29% | 1 | 1 | 0% | 1,017 | 1,214 | +19% | 0 | 0 | — |
case-21 | pass→pass | 7,719 | 4,556 | -41% | 1 | 1 | 0% | 1,304 | 1,094 | -16% | 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 21 counted toward the lift figure. The other 1 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 +23 percentage points is the difference between those two pass rates over the 21 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.