Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Shared Python package for Science Skills, currently containing http_client -- a unified HTTP client with rate limiting, retries, and exponential backoff. Not a standalone agent skill. Do not invoke directly.
.claude/skills/mkurman-scienceskillscommon/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -30% | 0% |
This is a shared Python package, not an agent skill. Skills import it as:
pythonfrom science_skills.scienceskillscommon import http_client
Each skill declares this as a dependency in its inline uv script header, so it is installed automatically on first use.
This SKILL.md file is included so that standard skill installers automatically discover and install this package alongside the skills that depend on it.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 6,408 | 4,410 | -31% | 1 | 1 | 0% | 1,277 | 907 | -29% | 0 | 0 | — |
case-02 | fail→pass | 17,197 | 3,627 | -79% | 1 | 1 | 0% | 1,425 | 708 | -50% | 0 | 0 | — |
case-03 | pass→pass | 12,399 | 10,534 | -15% | 1 | 1 | 0% | 2,443 | 2,396 | -2% | 0 | 0 | — |
case-04 | pass→pass | 8,148 | 6,779 | -17% | 1 | 1 | 0% | 1,649 | 1,460 | -11% | 0 | 0 | — |
case-05 | pass→pass | 7,154 | 3,834 | -46% | 1 | 1 | 0% | 1,283 | 854 | -33% | 0 | 0 | — |
case-06 | fail→pass | 16,551 | 2,482 | -85% | 1 | 1 | 0% | 1,146 | 513 | -55% | 0 | 0 | — |
case-07 | fail→pass | 8,310 | 3,791 | -54% | 1 | 1 | 0% | 1,528 | 755 | -51% | 0 | 0 | — |
case-08 | fail→pass | 13,831 | 8,051 | -42% | 1 | 1 | 0% | 2,535 | 1,764 | -30% | 0 | 0 | — |
case-09 | fail→pass | 12,326 | 2,299 | -81% | 1 | 1 | 0% | 1,928 | 454 | -76% | 0 | 0 | — |
case-10 | pass→pass | 7,797 | 2,821 | -64% | 1 | 1 | 0% | 1,152 | 528 | -54% | 0 | 0 | — |
case-11 | fail→pass | 6,446 | 3,593 | -44% | 1 | 1 | 0% | 1,126 | 704 | -37% | 0 | 0 | — |
case-12 | fail→pass | 10,688 | 5,323 | -50% | 1 | 1 | 0% | 1,878 | 1,162 | -38% | 0 | 0 | — |
case-13 | fail→pass | 3,573 | 3,177 | -11% | 1 | 1 | 0% | 593 | 577 | -3% | 0 | 0 | — |
case-14 | fail→pass | 17,169 | 3,425 | -80% | 1 | 1 | 0% | 1,688 | 658 | -61% | 0 | 0 | — |
case-15 | fail→pass | 15,158 | 3,427 | -77% | 1 | 1 | 0% | 2,275 | 642 | -72% | 0 | 0 | — |
case-16 | fail→pass | 6,683 | 1,544 | -77% | 1 | 1 | 0% | 1,044 | 311 | -70% | 0 | 0 | — |
case-17 | pass→pass | 12,903 | 6,614 | -49% | 1 | 1 | 0% | 1,987 | 1,175 | -41% | 0 | 0 | — |
case-18 | fail→pass | 9,705 | 5,764 | -41% | 1 | 1 | 0% | 1,791 | 1,063 | -41% | 0 | 0 | — |
case-19 | fail→pass | 8,648 | 3,135 | -64% | 1 | 1 | 0% | 1,303 | 550 | -58% | 0 | 0 | — |
case-20 | fail→pass | 13,711 | 2,761 | -80% | 1 | 1 | 0% | 1,294 | 539 | -58% | 0 | 0 | — |
case-21 | fail→pass | 6,944 | 3,865 | -44% | 1 | 1 | 0% | 1,203 | 718 | -40% | 0 | 0 | — |
case-22 | fail→pass | 6,173 | 3,332 | -46% | 1 | 1 | 0% | 966 | 662 | -31% | 0 | 0 | — |
case-23 | fail→pass | 24,606 | 5,944 | -76% | 1 | 1 | 0% | 2,150 | 1,168 | -46% | 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. 23 cases were attempted, and 22 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 +67 percentage points is the difference between those two pass rates over the 22 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.