Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Standards-compliant code for the Research Project Template — infrastructure or project src/, thin orchestrators, type hints, no mocks. USE WHEN implementing an algorithm, utility, analysis method, optimizer, new module in projects/*/src or infrastructure/, or user says add code following template architecture — even without docs/prompts. Not for end-to-end feature spanning manuscript + pipeline (use template-feature-addition).
.claude/skills/docxology-template-code-development/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -23% | 0% |
infrastructure (generic, 60% cov) vs project (domain, 90% cov).docs/_generated/active_projects.md.infrastructure/ or projects/<n>/src/; scripts only orchestrate I/O and plotting.get_logger(__name__); custom exceptions extending TemplateError; Google docstrings.docs/rules/testing_standards.md).bashuv run python -m infrastructure.project.public_scope lint-paths | xargs uv run ruff check uv run python -m infrastructure.project.public_scope source-paths | xargs uv run mypy uv run pytest projects/<project>/tests/ --cov=projects/<project>/src --cov-fail-under=90 -q # or infra: uv run pytest tests/infra_tests/ --cov=infrastructure --cov-fail-under=60 -q
docs/rules/ — normative standardsdocs/core/architecture.md — two layers, thin orchestrator| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | fail→fail | 13,003 | 7,949 | -39% | 1 | 1 | 0% | 2,367 | 2,054 | -13% | 0 | 0 | — |
case-01 | fail→fail | 5,633 | 5,925 | +5% | 1 | 1 | 0% | 142 | 832 | +486% | 0 | 0 | — |
case-02 | fail→fail | 3,370 | 4,531 | +34% | 1 | 1 | 0% | 143 | 807 | +464% | 0 | 0 | — |
case-03 | fail→fail | 3,414 | 4,821 | +41% | 1 | 1 | 0% | 223 | 786 | +252% | 0 | 0 | — |
case-04 | fail→pass | 19,512 | 16,114 | -17% | 1 | 1 | 0% | 4,813 | 4,467 | -7% | 0 | 0 | — |
case-05 | fail→pass | 9,209 | 3,276 | -64% | 1 | 1 | 0% | 1,606 | 1,137 | -29% | 0 | 0 | — |
case-06 | pass→pass | 7,394 | 16,017 | +117% | 1 | 1 | 0% | 1,248 | 3,337 | +167% | 0 | 0 | — |
case-07 | fail→pass | 6,846 | 1,414 | -79% | 1 | 1 | 0% | 1,190 | 763 | -36% | 0 | 0 | — |
case-08 | pass→fail | 14,130 | 2,488 | -82% | 1 | 1 | 0% | 3,047 | 933 | -69% | 0 | 0 | — |
case-09 | pass→pass | 11,912 | 5,079 | -57% | 1 | 1 | 0% | 2,111 | 1,493 | -29% | 0 | 0 | — |
case-10 | fail→fail | 7,097 | 2,742 | -61% | 1 | 1 | 0% | 1,194 | 1,013 | -15% | 0 | 0 | — |
case-11 | pass→pass | 5,436 | 4,592 | -16% | 1 | 1 | 0% | 1,045 | 1,338 | +28% | 0 | 0 | — |
case-12 | pass→pass | 9,420 | 3,580 | -62% | 1 | 1 | 0% | 1,827 | 1,255 | -31% | 0 | 0 | — |
case-13 | pass→pass | 14,107 | 3,982 | -72% | 1 | 1 | 0% | 2,265 | 1,244 | -45% | 0 | 0 | — |
case-14 | fail→pass | 9,973 | 2,242 | -78% | 1 | 1 | 0% | 1,868 | 890 | -52% | 0 | 0 | — |
case-15 | fail→fail | 11,825 | 2,943 | -75% | 1 | 1 | 0% | 1,961 | 919 | -53% | 0 | 0 | — |
case-16 | pass→pass | 10,867 | 2,122 | -80% | 1 | 1 | 0% | 1,830 | 897 | -51% | 0 | 0 | — |
case-17 | pass→pass | 7,964 | 3,893 | -51% | 1 | 1 | 0% | 1,402 | 1,261 | -10% | 0 | 0 | — |
case-18 | fail→pass | 5,633 | 1,454 | -74% | 1 | 1 | 0% | 1,022 | 790 | -23% | 0 | 0 | — |
case-19 | fail→pass | 10,235 | 2,181 | -79% | 1 | 1 | 0% | 1,398 | 851 | -39% | 0 | 0 | — |
case-21 | pass→pass | 9,332 | 3,719 | -60% | 1 | 1 | 0% | 1,563 | 1,210 | -23% | 0 | 0 | — |
case-22 | fail→pass | 14,343 | 3,156 | -78% | 1 | 1 | 0% | 2,192 | 1,039 | -53% | 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 19 counted toward the lift figure. The other 3 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 19 comparable cases. 2 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.