Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Research-to-publication orchestrator for template projects: research, write, verify, review, revise, reproduce, validate, and finalize. USE WHEN the user wants the whole paper workflow or enters midstream with an existing paper or reviewer comments.
.claude/skills/docxology-template-academic-pipeline/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 4% | 0% |
Template-native orchestrator for the full research-to-publication path. It coordinates existing skills and file-backed controls; it does not introduce an autonomous hidden approval loop.
output/hitl/.output/ as the fix.bashuv run python scripts/runner/execute_pipeline.py --project <project> --core-only uv run python -m infrastructure.validation.cli evidence projects/<project> --fail-on-issues uv run python -m infrastructure.reference.verification verify projects/<project>/manuscript/references.bib --live --as-of-year <year> --fail-on-issues uv run python -m infrastructure.validation.cli prose-quality projects/<project>/manuscript uv run python -m infrastructure.validation.cli integrity output/<project>/ uv run python -m infrastructure.core.pipeline.snapshot compare <left> <right> --output-dir projects/<project>/output
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | fail→pass | 7,390 | 4,802 | -35% | 1 | 1 | 0% | 1,178 | 1,530 | +30% | 0 | 0 | — |
case-01 | fail→fail | 4,524 | 20,448 | +352% | 1 | 1 | 0% | 289 | 4,340 | +1402% | 0 | 0 | — |
case-02 | fail→fail | 18,045 | 4,507 | -75% | 1 | 1 | 0% | 3,188 | 918 | -71% | 0 | 0 | — |
case-03 | fail→fail | 4,284 | 7,051 | +65% | 1 | 1 | 0% | 241 | 1,150 | +377% | 0 | 0 | — |
case-04 | fail→pass | 14,722 | 20,105 | +37% | 1 | 1 | 0% | 2,010 | 2,622 | +30% | 0 | 0 | — |
case-05 | pass→pass | 8,733 | 4,467 | -49% | 1 | 1 | 0% | 1,482 | 1,454 | -2% | 0 | 0 | — |
case-06 | pass→pass | 19,506 | 18,256 | -6% | 1 | 1 | 0% | 4,026 | 4,311 | +7% | 0 | 0 | — |
case-08 | fail→pass | 12,785 | 7,426 | -42% | 1 | 1 | 0% | 2,219 | 2,029 | -9% | 0 | 0 | — |
case-09 | fail→pass | 14,718 | 3,357 | -77% | 1 | 1 | 0% | 2,564 | 1,368 | -47% | 0 | 0 | — |
case-10 | fail→pass | 4,964 | 1,890 | -62% | 1 | 1 | 0% | 979 | 1,017 | +4% | 0 | 0 | — |
case-11 | fail→pass | 11,312 | 2,289 | -80% | 1 | 1 | 0% | 1,969 | 1,116 | -43% | 0 | 0 | — |
case-12 | fail→pass | 7,571 | 3,036 | -60% | 1 | 1 | 0% | 1,274 | 1,323 | +4% | 0 | 0 | — |
case-13 | fail→pass | 12,277 | 2,023 | -84% | 1 | 1 | 0% | 2,115 | 1,062 | -50% | 0 | 0 | — |
case-14 | fail→pass | 6,482 | 2,957 | -54% | 1 | 1 | 0% | 1,009 | 1,215 | +20% | 0 | 0 | — |
case-15 | fail→pass | 14,328 | 3,386 | -76% | 1 | 1 | 0% | 2,330 | 1,264 | -46% | 0 | 0 | — |
case-16 | fail→pass | 8,105 | 2,246 | -72% | 1 | 1 | 0% | 1,510 | 1,036 | -31% | 0 | 0 | — |
case-17 | fail→pass | 10,706 | 1,397 | -87% | 1 | 1 | 0% | 1,766 | 902 | -49% | 0 | 0 | — |
case-18 | fail→fail | 7,466 | 3,896 | -48% | 1 | 1 | 0% | 1,196 | 1,546 | +29% | 0 | 0 | — |
case-19 | fail→pass | 11,167 | 4,944 | -56% | 1 | 1 | 0% | 1,796 | 1,660 | -8% | 0 | 0 | — |
case-20 | fail→pass | 20,660 | 9,063 | -56% | 1 | 1 | 0% | 1,914 | 2,384 | +25% | 0 | 0 | — |
case-21 | fail→pass | 12,709 | 1,992 | -84% | 1 | 1 | 0% | 1,811 | 1,064 | -41% | 0 | 0 | — |
case-22 | pass→pass | 6,610 | 2,590 | -61% | 1 | 1 | 0% | 925 | 1,137 | +23% | 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 +68 percentage points is the difference between those two pass rates over the 19 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.