Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Multi-phase literature-review exemplar with phase-aware retrieval, filtering, provenance, and cross-phase validation.
.claude/skills/docxology-template-advanced-literature-review/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 58 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -57% | 0% |
Load this project-scoped skill when working inside projects/templates/template_advanced_literature_review/ or forking it into a new phased literature-review pipeline.
bashuv run python scripts/pipeline/stage_01_test.py --project templates/template_advanced_literature_review --project-only uv run python scripts/pipeline/stage_02_analysis.py --project templates/template_advanced_literature_review uv run python scripts/pipeline/stage_03_render.py --project templates/template_advanced_literature_review uv run python scripts/pipeline/stage_04_validate.py --project templates/template_advanced_literature_review
src/multi_phase/ and numbered scripts thin.Prefer MCP connectors and configured retrieval keys first. Monid (infrastructure/search/monid/) is infrastructure-only gap filler — see PRICING.md. Offline deep_research replay remains the exemplar default for long-form synthesis.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 49,756 | 15,555 | -69% | 1 | 1 | 0% | 8,243 | 669 | -92% | 0 | 0 | — |
case-02 | fail→fail | 23,044 | 16,501 | -28% | 1 | 1 | 0% | 2,645 | 714 | -73% | 0 | 0 | — |
case-03 | fail→fail | 13,562 | 16,304 | +20% | 1 | 1 | 0% | 172 | 609 | +254% | 0 | 0 | — |
case-04 | fail→pass | 12,281 | 9,501 | -23% | 1 | 1 | 0% | 956 | 993 | +4% | 0 | 0 | — |
case-05 | fail→pass | 12,957 | 9,964 | -23% | 1 | 1 | 0% | 1,054 | 1,216 | +15% | 0 | 0 | — |
case-06 | fail→pass | 12,992 | 11,089 | -15% | 1 | 1 | 0% | 1,216 | 1,252 | +3% | 0 | 0 | — |
case-07 | fail→pass | 13,209 | 7,138 | -46% | 1 | 1 | 0% | 1,422 | 819 | -42% | 0 | 0 | — |
case-13 | fail→pass | 12,839 | 7,845 | -39% | 1 | 1 | 0% | 1,884 | 816 | -57% | 0 | 0 | — |
case-08 | fail→pass | 15,778 | 7,504 | -52% | 1 | 1 | 0% | 1,772 | 862 | -51% | 0 | 0 | — |
case-09 | fail→pass | 18,847 | 9,257 | -51% | 1 | 1 | 0% | 2,433 | 1,088 | -55% | 0 | 0 | — |
case-10 | fail→pass | 19,418 | 11,200 | -42% | 1 | 1 | 0% | 2,071 | 1,447 | -30% | 0 | 0 | — |
case-11 | fail→pass | 9,939 | 3,539 | -64% | 1 | 1 | 0% | 1,497 | 928 | -38% | 0 | 0 | — |
case-12 | fail→pass | 15,363 | 8,165 | -47% | 1 | 1 | 0% | 1,651 | 877 | -47% | 0 | 0 | — |
case-14 | pass→pass | 14,351 | 1,816 | -87% | 1 | 1 | 0% | 2,126 | 630 | -70% | 0 | 0 | — |
case-15 | pass→pass | 7,648 | 5,538 | -28% | 1 | 1 | 0% | 1,077 | 1,095 | +2% | 0 | 0 | — |
case-16 | pass→pass | 13,037 | 2,658 | -80% | 1 | 1 | 0% | 1,363 | 704 | -48% | 0 | 0 | — |
case-17 | fail→pass | 11,717 | 3,244 | -72% | 1 | 1 | 0% | 1,110 | 816 | -26% | 0 | 0 | — |
case-18 | pass→pass | 25,964 | 9,565 | -63% | 1 | 1 | 0% | 2,646 | 1,045 | -61% | 0 | 0 | — |
case-19 | fail→fail | 20,990 | 15,473 | -26% | 1 | 1 | 0% | 2,423 | 2,024 | -16% | 0 | 0 | — |
case-20 | pass→pass | 18,314 | 19,504 | +6% | 1 | 1 | 0% | 2,163 | 2,092 | -3% | 0 | 0 | — |
case-21 | fail→fail | 14,018 | 17,252 | +23% | 1 | 1 | 0% | 2,794 | 3,703 | +33% | 0 | 0 | — |
case-22 | pass→pass | 8,537 | 6,021 | -29% | 1 | 1 | 0% | 654 | 1,119 | +71% | 0 | 0 | — |
case-23 | pass→pass | 13,801 | 12,615 | -9% | 1 | 1 | 0% | 1,715 | 1,945 | +13% | 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, and 20 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 +48 percentage points is the difference between those two pass rates over the 20 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/4/2026 | +64% |
Other measured skills in the registry, with their headline benchmark lift.