Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →AutoResearch loop exemplar — deterministic ML candidate evaluation, evidence registries, claim ledgers, artifact manifests, readiness gates.
.claude/skills/docxology-template-autoresearch-project/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 285% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -37% | 0% |
Project-scoped skill for the in-repo exemplar at projects/templates/template_autoresearch_project/. Load this when working inside the project.
template_autoresearch_project 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_autoresearch_project/tests --cov=projects/templates/template_autoresearch_project/src --cov-fail-under=90 uv run python scripts/pipeline/stage_02_analysis.py --project templates/template_autoresearch_project uv run python scripts/pipeline/stage_03_render.py --project templates/template_autoresearch_project uv run python scripts/pipeline/stage_04_validate.py --project templates/template_autoresearch_project uv run python scripts/pipeline/stage_05_copy.py --project templates/template_autoresearch_project
src/ or sharedinfrastructure/, not in scripts/.
computation.
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-05 | fail→pass | 12,567 | 4,748 | -62% | 1 | 1 | 0% | 2,191 | 1,200 | -45% | 0 | 0 | — |
case-06 | fail→pass | 7,749 | 4,595 | -41% | 1 | 1 | 0% | 1,291 | 1,165 | -10% | 0 | 0 | — |
case-01 | fail→pass | 5,802 | 2,284 | -61% | 1 | 1 | 0% | 940 | 816 | -13% | 0 | 0 | — |
case-02 | fail→fail | 10,905 | 3,497 | -68% | 1 | 1 | 0% | 1,868 | 882 | -53% | 0 | 0 | — |
case-03 | fail→pass | 2,418 | 3,257 | +35% | 1 | 1 | 0% | 280 | 1,079 | +285% | 0 | 0 | — |
case-04 | pass→pass | 10,808 | 4,825 | -55% | 1 | 1 | 0% | 1,909 | 1,363 | -29% | 0 | 0 | — |
case-07 | fail→pass | 8,837 | 3,031 | -66% | 1 | 1 | 0% | 1,641 | 1,036 | -37% | 0 | 0 | — |
case-08 | fail→pass | 7,858 | 1,394 | -82% | 1 | 1 | 0% | 1,509 | 723 | -52% | 0 | 0 | — |
case-09 | fail→pass | 7,315 | 1,529 | -79% | 1 | 1 | 0% | 1,293 | 676 | -48% | 0 | 0 | — |
case-10 | fail→pass | 10,183 | 1,638 | -84% | 1 | 1 | 0% | 1,630 | 661 | -59% | 0 | 0 | — |
case-11 | fail→pass | 8,055 | 1,928 | -76% | 1 | 1 | 0% | 1,395 | 805 | -42% | 0 | 0 | — |
case-12 | fail→pass | 7,115 | 1,627 | -77% | 1 | 1 | 0% | 1,310 | 739 | -44% | 0 | 0 | — |
case-13 | fail→pass | 5,678 | 2,189 | -61% | 1 | 1 | 0% | 920 | 841 | -9% | 0 | 0 | — |
case-14 | fail→pass | 15,629 | 15,368 | -2% | 1 | 1 | 0% | 2,748 | 3,383 | +23% | 0 | 0 | — |
case-15 | pass→pass | 11,573 | 8,328 | -28% | 1 | 1 | 0% | 1,884 | 1,917 | +2% | 0 | 0 | — |
case-16 | pass→pass | 5,526 | 1,963 | -64% | 1 | 1 | 0% | 870 | 791 | -9% | 0 | 0 | — |
case-17 | pass→pass | 7,834 | 4,096 | -48% | 1 | 1 | 0% | 1,465 | 1,050 | -28% | 0 | 0 | — |
case-18 | fail→pass | 11,947 | 4,159 | -65% | 1 | 1 | 0% | 2,034 | 1,124 | -45% | 0 | 0 | — |
case-19 | pass→pass | 7,632 | 4,434 | -42% | 1 | 1 | 0% | 1,433 | 1,123 | -22% | 0 | 0 | — |
case-20 | fail→pass | 3,691 | 2,513 | -32% | 1 | 1 | 0% | 687 | 863 | +26% | 0 | 0 | — |
case-21 | fail→pass | 2,702 | 2,910 | +8% | 1 | 1 | 0% | 471 | 870 | +85% | 0 | 0 | — |
case-22 | pass→pass | 5,910 | 3,477 | -41% | 1 | 1 | 0% | 1,027 | 1,129 | +10% | 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 +68 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.