Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Fixture download scripts for test data. Downloads real GitHub repos (requests, fastapi) via sparse checkout and M4 monthly timeseries data from Hugging Face with synthetic fallback. Resumable — skips if target directory already exists and is non-empty.
.claude/skills/docxology-template-fixtures/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 10 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -72% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -70% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -67% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -68% | 0% |
Test fixture download scripts.
bashuv run python scripts/fixtures/download_real_codebases.py uv run python scripts/fixtures/download_timeseries_benchmarks.py
scripts/fixtures/AGENTS.md — detailed documentationtests/fixtures/ — fixture storage directory| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 2,477 | 1,474 | -40% | 1 | 1 | 0% | 322 | 273 | -15% | 0 | 0 | — |
case-02 | fail→pass | 6,972 | 1,806 | -74% | 1 | 1 | 0% | 1,196 | 339 | -72% | 0 | 0 | — |
case-03 | fail→pass | 7,215 | 1,771 | -75% | 1 | 1 | 0% | 1,166 | 346 | -70% | 0 | 0 | — |
case-04 | fail→pass | 6,272 | 1,852 | -70% | 1 | 1 | 0% | 1,186 | 387 | -67% | 0 | 0 | — |
case-05 | fail→pass | 7,813 | 2,241 | -71% | 1 | 1 | 0% | 1,398 | 446 | -68% | 0 | 0 | — |
case-06 | fail→pass | 8,297 | 3,770 | -55% | 1 | 1 | 0% | 1,462 | 737 | -50% | 0 | 0 | — |
case-07 | pass→pass | 2,870 | 1,811 | -37% | 1 | 1 | 0% | 466 | 336 | -28% | 0 | 0 | — |
case-08 | fail→pass | 3,611 | 1,397 | -61% | 1 | 1 | 0% | 610 | 278 | -54% | 0 | 0 | — |
case-09 | fail→pass | 4,905 | 1,632 | -67% | 1 | 1 | 0% | 764 | 340 | -55% | 0 | 0 | — |
case-10 | fail→pass | 4,033 | 2,420 | -40% | 1 | 1 | 0% | 606 | 349 | -42% | 0 | 0 | — |
case-11 | pass→pass | 7,667 | 1,956 | -74% | 1 | 1 | 0% | 1,453 | 414 | -72% | 0 | 0 | — |
case-12 | fail→pass | 4,073 | 3,347 | -18% | 1 | 1 | 0% | 578 | 658 | +14% | 0 | 0 | — |
case-13 | fail→pass | 6,453 | 2,072 | -68% | 1 | 1 | 0% | 957 | 406 | -58% | 0 | 0 | — |
case-14 | pass→pass | 8,915 | 1,910 | -79% | 1 | 1 | 0% | 1,529 | 354 | -77% | 0 | 0 | — |
case-15 | fail→pass | 7,095 | 2,075 | -71% | 1 | 1 | 0% | 1,257 | 409 | -67% | 0 | 0 | — |
case-16 | fail→pass | 13,117 | 1,755 | -87% | 1 | 1 | 0% | 1,666 | 373 | -78% | 0 | 0 | — |
case-17 | fail→pass | 7,676 | 1,797 | -77% | 1 | 1 | 0% | 1,410 | 343 | -76% | 0 | 0 | — |
case-18 | pass→pass | 4,088 | 2,163 | -47% | 1 | 1 | 0% | 626 | 343 | -45% | 0 | 0 | — |
case-19 | pass→pass | 11,282 | 2,004 | -82% | 1 | 1 | 0% | 1,892 | 374 | -80% | 0 | 0 | — |
case-20 | fail→pass | 10,197 | 5,844 | -43% | 1 | 1 | 0% | 1,675 | 877 | -48% | 0 | 0 | — |
case-21 | pass→pass | 8,307 | 4,129 | -50% | 1 | 1 | 0% | 1,266 | 754 | -40% | 0 | 0 | — |
case-22 | fail→fail | 8,881 | 5,773 | -35% | 1 | 1 | 0% | 1,578 | 1,093 | -31% | 0 | 0 | — |
case-23 | fail→pass | 5,260 | 2,353 | -55% | 1 | 1 | 0% | 972 | 379 | -61% | 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. The headline lift of +70 percentage points is the difference between those two pass rates over the 23 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.