Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Core orchestration guidelines and MCP interactions for the src/ library.
.claude/skills/docxology-meta-analysis-source-api/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -82% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -70% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -50% | 0% |
You are interfacing with the src/ directory of the literature meta-analysis project. This directory contains 45+ public APIs spread across 6 submodules.
When operating within this workspace, adhere to the following interaction protocols:
pytest-httpserver or local data objects. Do not use mocker.patch or MagicMock.scripts/, or by running uv run pytest in the tests/ directory. Do not write temporary execution blocks inside src/.manuscript/config.yaml using your file reading tools to understand runtime constraints.scripts/. It belongs here.seed=42) to guarantee deterministic analysis.python3 -m infrastructure.validation.cli markdown to verify nothing was broken.Refer to the specific SKILL.md in each subdirectory for granular file-level guidance.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 17,003 | 6,119 | -64% | 1 | 1 | 0% | 3,790 | 631 | -83% | 0 | 0 | — |
case-02 | fail→fail | 3,495 | 5,215 | +49% | 1 | 1 | 0% | 167 | 609 | +265% | 0 | 0 | — |
case-03 | fail→fail | 3,469 | 5,947 | +71% | 1 | 1 | 0% | 177 | 645 | +264% | 0 | 0 | — |
case-19 | pass→fail | 16,947 | 4,927 | -71% | 1 | 1 | 0% | 3,010 | 570 | -81% | 0 | 0 | — |
case-04 | fail→fail | 13,157 | 6,652 | -49% | 1 | 1 | 0% | 2,846 | 665 | -77% | 0 | 0 | — |
case-05 | fail→pass | 14,585 | 9,478 | -35% | 1 | 1 | 0% | 2,824 | 2,118 | -25% | 0 | 0 | — |
case-06 | fail→pass | 4,908 | 2,850 | -42% | 1 | 1 | 0% | 817 | 734 | -10% | 0 | 0 | — |
case-07 | pass→pass | 5,381 | 2,908 | -46% | 1 | 1 | 0% | 931 | 787 | -15% | 0 | 0 | — |
case-08 | fail→pass | 20,173 | 1,817 | -91% | 1 | 1 | 0% | 3,342 | 606 | -82% | 0 | 0 | — |
case-09 | fail→pass | 13,782 | 2,856 | -79% | 1 | 1 | 0% | 2,519 | 758 | -70% | 0 | 0 | — |
case-10 | pass→pass | 11,978 | 3,898 | -67% | 1 | 1 | 0% | 2,086 | 968 | -54% | 0 | 0 | — |
case-11 | fail→fail | 15,406 | 7,344 | -52% | 1 | 1 | 0% | 3,044 | 894 | -71% | 0 | 0 | — |
case-12 | fail→fail | 12,027 | 5,161 | -57% | 1 | 1 | 0% | 2,341 | 1,327 | -43% | 0 | 0 | — |
case-13 | pass→pass | 3,531 | 2,000 | -43% | 1 | 1 | 0% | 667 | 596 | -11% | 0 | 0 | — |
case-14 | fail→pass | 10,942 | 8,061 | -26% | 1 | 1 | 0% | 2,355 | 1,177 | -50% | 0 | 0 | — |
case-15 | fail→pass | 19,265 | 2,037 | -89% | 1 | 1 | 0% | 1,018 | 635 | -38% | 0 | 0 | — |
case-16 | fail→fail | 15,764 | 4,815 | -69% | 1 | 1 | 0% | 2,949 | 513 | -83% | 0 | 0 | — |
case-17 | pass→pass | 11,309 | 5,034 | -55% | 1 | 1 | 0% | 2,052 | 1,287 | -37% | 0 | 0 | — |
case-18 | fail→pass | 10,217 | 9,164 | -10% | 1 | 1 | 0% | 2,033 | 1,868 | -8% | 0 | 0 | — |
case-20 | pass→fail | 10,573 | 4,726 | -55% | 1 | 1 | 0% | 1,997 | 547 | -73% | 0 | 0 | — |
case-21 | pass→pass | 28,579 | 18,289 | -36% | 1 | 1 | 0% | 2,198 | 3,914 | +78% | 0 | 0 | — |
case-22 | fail→pass | 8,716 | 3,356 | -61% | 1 | 1 | 0% | 1,562 | 869 | -44% | 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 14 counted toward the lift figure. The other 8 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 +27 percentage points is the difference between those two pass rates over the 14 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.
Other measured skills in the registry, with their headline benchmark lift.