Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Deterministic token-injection manuscript generator — config-owned lexicon, conditional IMRAD, QA probes, authoring contract, and token provenance.
.claude/skills/docxology-template-madlib/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -15% | 0% |
Project-scoped skill for the in-repo exemplar at projects/templates/template_madlib/. Load this when working inside the project.
template_madlib exemplar — running scripts, editing source,or regenerating outputs.
no-mocks testing) still hold after changes.
bash# From the repository root uv run pytest projects/templates/template_madlib/tests --cov=projects/templates/template_madlib/src --cov-fail-under=90 uv run python scripts/pipeline/stage_02_analysis.py --project templates/template_madlib uv run python scripts/pipeline/stage_03_render.py --project templates/template_madlib uv run python scripts/pipeline/stage_04_validate.py --project templates/template_madlib uv run python scripts/pipeline/stage_05_copy.py --project templates/template_madlib
src/ modules (run.py, composition_*, figure_specs, analysis_*), not in scripts/.composition.py re-exports in tests; edit split modules for body/table/figure changes.output/ — regenerate fromsource and config.
as working directory unless the child AGENTS.md states otherwise.
AGENTS.mdREADME.mdTODO.mdprojects/AGENTS.md| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | fail→pass | 11,762 | 4,743 | -60% | 1 | 1 | 0% | 2,253 | 1,270 | -44% | 0 | 0 | — |
case-06 | fail→pass | 9,908 | 2,837 | -71% | 1 | 1 | 0% | 1,664 | 1,008 | -39% | 0 | 0 | — |
case-01 | fail→fail | 3,056 | 2,019 | -34% | 1 | 1 | 0% | 338 | 686 | +103% | 0 | 0 | — |
case-02 | fail→fail | 18,906 | 3,140 | -83% | 1 | 1 | 0% | 2,497 | 870 | -65% | 0 | 0 | — |
case-03 | fail→fail | 5,220 | 2,982 | -43% | 1 | 1 | 0% | 865 | 752 | -13% | 0 | 0 | — |
case-04 | fail→pass | 15,497 | 7,207 | -53% | 1 | 1 | 0% | 2,875 | 1,851 | -36% | 0 | 0 | — |
case-05 | fail→pass | 12,922 | 7,215 | -44% | 1 | 1 | 0% | 2,200 | 1,472 | -33% | 0 | 0 | — |
case-07 | fail→pass | 12,617 | 4,138 | -67% | 1 | 1 | 0% | 1,449 | 1,230 | -15% | 0 | 0 | — |
case-08 | fail→pass | 7,113 | 3,735 | -47% | 1 | 1 | 0% | 1,313 | 1,193 | -9% | 0 | 0 | — |
case-09 | pass→pass | 7,141 | 3,118 | -56% | 1 | 1 | 0% | 1,191 | 1,071 | -10% | 0 | 0 | — |
case-10 | fail→pass | 13,002 | 8,257 | -36% | 1 | 1 | 0% | 2,091 | 2,304 | +10% | 0 | 0 | — |
case-11 | fail→pass | 14,482 | 7,646 | -47% | 1 | 1 | 0% | 2,341 | 1,641 | -30% | 0 | 0 | — |
case-13 | fail→pass | 8,983 | 2,673 | -70% | 1 | 1 | 0% | 1,376 | 1,021 | -26% | 0 | 0 | — |
case-14 | pass→pass | 9,306 | 3,262 | -65% | 1 | 1 | 0% | 1,412 | 1,056 | -25% | 0 | 0 | — |
case-15 | pass→pass | 10,382 | 3,048 | -71% | 1 | 1 | 0% | 1,897 | 1,054 | -44% | 0 | 0 | — |
case-16 | pass→pass | 9,540 | 3,250 | -66% | 1 | 1 | 0% | 1,869 | 994 | -47% | 0 | 0 | — |
case-17 | pass→pass | 8,623 | 3,552 | -59% | 1 | 1 | 0% | 1,459 | 1,178 | -19% | 0 | 0 | — |
case-18 | pass→pass | 8,585 | 4,532 | -47% | 1 | 1 | 0% | 1,703 | 1,278 | -25% | 0 | 0 | — |
case-19 | pass→pass | 11,292 | 7,667 | -32% | 1 | 1 | 0% | 2,284 | 1,873 | -18% | 0 | 0 | — |
case-20 | pass→pass | 4,595 | 2,749 | -40% | 1 | 1 | 0% | 812 | 1,033 | +27% | 0 | 0 | — |
case-21 | pass→pass | 8,764 | 6,577 | -25% | 1 | 1 | 0% | 1,391 | 1,384 | -1% | 0 | 0 | — |
case-22 | pass→pass | 6,361 | 6,379 | +0% | 1 | 1 | 0% | 1,096 | 1,301 | +19% | 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. The headline lift of +41 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.