Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate a commit message, commit all changes, run deploy_pip.ps1 to publish to PyPI, and tag HEAD with the deployed version on success
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-18 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-19 | ✓→✗ | ▼ Worse | -43% | 0% |
| case-12 | ✓→✗ | ▼ Worse | -24% | 0% |
When instructed to deploy, release, or publish to PyPI, follow these steps in order. GitHub Actions builds the tagged commit, publishes it with the repository's PYPI_API_TOKEN secret, and creates the GitHub Release.
// turbo
powershellgit status --short
If there are uncommitted changes:
git diff --stat and git diff to review them.// turbo
powershellgit diff --stat
// turbo
powershellgit diff
// turbo powershell git add -A // turbo powershell git commit -m "<generated commit message>" If there are NO uncommitted changes, proceed directly to Step 2.
Choose the exact package version to release (for example, 1.0.98). The Git tag must be v<version>, and the workflow passes <version> to the build through OK_SCRIPT_BUILD_VERSION.
// turbo
powershellgit tag v<version>
Replace <version> with the chosen version (e.g. 1.0.98).
// turbo
powershellgit push --tags
Open the GitHub Actions run triggered by the pushed tag. It must complete both the Build and publish and Create GitHub release jobs. If publishing fails, do not create or move another tag; report the failure.
Tell the user:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-18 | fail→pass | 5,633 | 1,841 | -67% | 1 | 1 | 0% | 1,020 | 765 | -25% | 0 | 0 | — |
case-01 | fail→fail | 2,443 | 4,825 | +98% | 1 | 1 | 0% | 314 | 690 | +120% | 0 | 0 | — |
case-02 | fail→fail | 4,142 | 3,479 | -16% | 1 | 1 | 0% | 187 | 576 | +208% | 0 | 0 | — |
case-03 | fail→fail | 10,145 | 4,780 | -53% | 1 | 1 | 0% | 1,322 | 557 | -58% | 0 | 0 | — |
case-19 | pass→fail | 7,716 | 2,227 | -71% | 1 | 1 | 0% | 1,415 | 807 | -43% | 0 | 0 | — |
case-04 | fail→pass | 2,800 | 1,946 | -31% | 1 | 1 | 0% | 385 | 667 | +73% | 0 | 0 | — |
case-05 | pass→pass | 5,525 | 1,577 | -71% | 1 | 1 | 0% | 1,065 | 732 | -31% | 0 | 0 | — |
case-06 | pass→pass | 5,479 | 1,713 | -69% | 1 | 1 | 0% | 1,034 | 650 | -37% | 0 | 0 | — |
case-07 | fail→fail | 8,646 | 2,566 | -70% | 1 | 1 | 0% | 1,362 | 591 | -57% | 0 | 0 | — |
case-08 | pass→pass | 9,053 | 2,111 | -77% | 1 | 1 | 0% | 1,629 | 830 | -49% | 0 | 0 | — |
case-09 | pass→pass | 1,835 | 1,377 | -25% | 1 | 1 | 0% | 233 | 658 | +182% | 0 | 0 | — |
case-10 | fail→fail | 11,110 | 4,553 | -59% | 1 | 1 | 0% | 1,787 | 1,188 | -34% | 0 | 0 | — |
case-11 | pass→pass | 9,948 | 1,949 | -80% | 1 | 1 | 0% | 1,602 | 756 | -53% | 0 | 0 | — |
case-12 | pass→fail | 5,508 | 1,735 | -69% | 1 | 1 | 0% | 975 | 740 | -24% | 0 | 0 | — |
case-13 | pass→pass | 11,791 | 3,056 | -74% | 1 | 1 | 0% | 1,953 | 818 | -58% | 0 | 0 | — |
case-14 | fail→fail | 10,855 | 2,391 | -78% | 1 | 1 | 0% | 1,862 | 857 | -54% | 0 | 0 | — |
case-15 | fail→pass | 8,104 | 2,019 | -75% | 1 | 1 | 0% | 1,401 | 712 | -49% | 0 | 0 | — |
case-16 | fail→fail | 3,263 | 4,043 | +24% | 1 | 1 | 0% | 496 | 560 | +13% | 0 | 0 | — |
case-17 | pass→pass | 4,665 | 1,295 | -72% | 1 | 1 | 0% | 756 | 596 | -21% | 0 | 0 | — |
case-20 | pass→pass | 5,170 | 3,320 | -36% | 1 | 1 | 0% | 935 | 1,001 | +7% | 0 | 0 | — |
case-21 | pass→pass | 11,048 | 7,450 | -33% | 1 | 1 | 0% | 2,095 | 1,875 | -11% | 0 | 0 | — |
case-22 | pass→pass | 18,085 | 4,684 | -74% | 1 | 1 | 0% | 1,340 | 1,395 | +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 +5 percentage points is the difference between those two pass rates over the 18 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/28/2026 | +14% |
Other measured skills in the registry, with their headline benchmark lift.