Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when a broad paper candidate pool needs deterministic deduplication and a stable core set. **Trigger**: dedupe, rank, core set, 去重, 排序, 精选论文, 核心集合. **Use when**: 检索后需要把广覆盖集合收敛成可管理的 core set（用于 taxonomy/outline/mapping）。 **Skip if**: 已经有人手工整理了稳定的 `papers/core_set.csv`（无需再次 churn）。 **Network**: none. **Guardrail**: 偏 deterministic；输出应可重复（稳定 paper_id、字段规范）。
.claude/skills/willoscar-dedupe-rank/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -44% | 0% |
papers/core_set.csv（无需再次 churn）。Turns a raw candidate pool into a deduped pool and a stable core set.
papers/papers_raw.jsonlpapers/papers_dedup.jsonlpapers/core_set.csvscripts/run.py should own only:
paper_id generationUse shared domain packs or pipeline contract metadata for topic-specific or product-specific behavior.
The script should prefer pipeline contract metadata over profile-name branching.
Current important field:
quality_contract.candidate_pool_policy.keep_full_deduped_poolquality_contract.candidate_pool_policy.include_domain_pinsquality_contract.candidate_pool_policy.minimum_domain_surveysquality_contract.candidate_pool_policy.survey_title_bonusIf true, the script keeps the full deduped pool in papers/core_set.csv unless the user explicitly overrides core size. Focused briefing Workflows can disable unconditional domain pins, reserve a smaller survey floor, and remove the generic survey-title bonus without changing broad Survey defaults.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-24 | pass→pass | 24,949 | 20,236 | -19% | 1 | 1 | 0% | 4,143 | 4,044 | -2% | 0 | 0 | — |
case-01 | fail→fail | 5,916 | 3,487 | -41% | 1 | 1 | 0% | 422 | 533 | +26% | 0 | 0 | — |
case-02 | fail→fail | 20,549 | 3,618 | -82% | 1 | 1 | 0% | 4,180 | 516 | -88% | 0 | 0 | — |
case-03 | fail→fail | 16,362 | 4,120 | -75% | 1 | 1 | 0% | 3,272 | 687 | -79% | 0 | 0 | — |
case-04 | fail→pass | 14,477 | 11,447 | -21% | 1 | 1 | 0% | 2,278 | 2,330 | +2% | 0 | 0 | — |
case-05 | fail→pass | 15,069 | 6,142 | -59% | 1 | 1 | 0% | 2,432 | 1,392 | -43% | 0 | 0 | — |
case-06 | pass→pass | 11,071 | 4,492 | -59% | 1 | 1 | 0% | 1,819 | 1,094 | -40% | 0 | 0 | — |
case-07 | fail→pass | 20,670 | 2,381 | -88% | 1 | 1 | 0% | 1,471 | 794 | -46% | 0 | 0 | — |
case-08 | fail→pass | 8,011 | 1,965 | -75% | 1 | 1 | 0% | 1,353 | 657 | -51% | 0 | 0 | — |
case-09 | fail→fail | 5,434 | 2,146 | -61% | 1 | 1 | 0% | 860 | 688 | -20% | 0 | 0 | — |
case-10 | fail→pass | 6,648 | 1,981 | -70% | 1 | 1 | 0% | 1,205 | 673 | -44% | 0 | 0 | — |
case-11 | pass→pass | 16,741 | 5,121 | -69% | 1 | 1 | 0% | 2,425 | 1,181 | -51% | 0 | 0 | — |
case-12 | pass→pass | 15,956 | 4,314 | -73% | 1 | 1 | 0% | 2,205 | 1,031 | -53% | 0 | 0 | — |
case-13 | pass→pass | 14,882 | 5,354 | -64% | 1 | 1 | 0% | 2,045 | 1,180 | -42% | 0 | 0 | — |
case-14 | fail→pass | 10,547 | 1,313 | -88% | 1 | 1 | 0% | 1,743 | 489 | -72% | 0 | 0 | — |
case-15 | fail→pass | 11,939 | 1,734 | -85% | 1 | 1 | 0% | 1,727 | 586 | -66% | 0 | 0 | — |
case-16 | pass→pass | 12,874 | 2,598 | -80% | 1 | 1 | 0% | 1,967 | 747 | -62% | 0 | 0 | — |
case-17 | pass→fail | 13,127 | 2,790 | -79% | 1 | 1 | 0% | 1,946 | 744 | -62% | 0 | 0 | — |
case-18 | pass→pass | 13,303 | 3,650 | -73% | 1 | 1 | 0% | 2,117 | 911 | -57% | 0 | 0 | — |
case-19 | pass→pass | 13,217 | 3,191 | -76% | 1 | 1 | 0% | 2,018 | 874 | -57% | 0 | 0 | — |
case-20 | pass→pass | 14,095 | 3,259 | -77% | 1 | 1 | 0% | 2,375 | 867 | -63% | 0 | 0 | — |
case-21 | fail→pass | 8,570 | 1,601 | -81% | 1 | 1 | 0% | 1,134 | 571 | -50% | 0 | 0 | — |
case-22 | pass→pass | 16,960 | 15,607 | -8% | 1 | 1 | 0% | 2,980 | 3,013 | +1% | 0 | 0 | — |
case-23 | pass→pass | 13,449 | 28,019 | +108% | 1 | 1 | 0% | 1,987 | 2,442 | +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. 24 cases were attempted, and 20 counted toward the lift figure. The other 4 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 +29 percentage points is the difference between those two pass rates over the 20 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.