Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Expand and refine search keywords (synonyms, acronyms, exclusions) and update `queries.md`. **Trigger**: keyword expansion, synonyms, exclusions, queries.md, 关键词扩展, 同义词, 排除词. **Use when**: 检索覆盖不足/噪声过大，或主题别名很多，需要系统化扩展与收敛检索词。 **Skip if**: `queries.md` 已经能稳定检出覆盖面（无需扩大范围导致后续成本爆炸）。 **Network**: none. **Guardrail**: 保持可控的 query 数量；明确 exclusions；避免“无限扩展”。
.claude/skills/willoscar-keyword-expansion/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-01 | ✓→✗ | ▼ Worse | 16% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | -14% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 53% | 0% |
queries.md without drifting scope.queries.mdDECISIONS.md scope notesqueries.mdqueries.md with a clear “why” note for each change.queries.md contains updated keywords and excludes.DECISIONS.md scope constraints.Fix:
DECISIONS.md scope notes and add explicit exclusions in queries.md for common false positives.Fix:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→fail | 13,508 | 12,795 | -5% | 1 | 1 | 0% | 2,365 | 2,752 | +16% | 0 | 0 | — |
case-02 | pass→pass | 14,834 | 11,184 | -25% | 1 | 1 | 0% | 2,175 | 1,881 | -14% | 0 | 0 | — |
case-03 | pass→pass | 12,107 | 18,263 | +51% | 1 | 1 | 0% | 2,162 | 3,303 | +53% | 0 | 0 | — |
case-12 | pass→pass | 12,976 | 10,134 | -22% | 1 | 1 | 0% | 1,802 | 1,900 | +5% | 0 | 0 | — |
case-04 | pass→pass | 20,140 | 18,856 | -6% | 1 | 1 | 0% | 4,079 | 4,434 | +9% | 0 | 0 | — |
case-05 | pass→pass | 9,677 | 7,696 | -20% | 1 | 1 | 0% | 1,504 | 1,548 | +3% | 0 | 0 | — |
case-06 | pass→pass | 6,148 | 7,758 | +26% | 1 | 1 | 0% | 980 | 1,496 | +53% | 0 | 0 | — |
case-07 | pass→pass | 5,121 | 6,952 | +36% | 1 | 1 | 0% | 797 | 1,380 | +73% | 0 | 0 | — |
case-08 | fail→pass | 6,714 | 6,422 | -4% | 1 | 1 | 0% | 1,075 | 1,313 | +22% | 0 | 0 | — |
case-09 | pass→pass | 4,315 | 9,133 | +112% | 1 | 1 | 0% | 697 | 1,184 | +70% | 0 | 0 | — |
case-10 | pass→pass | 8,333 | 5,997 | -28% | 1 | 1 | 0% | 1,293 | 1,192 | -8% | 0 | 0 | — |
case-11 | pass→pass | 13,493 | 8,806 | -35% | 1 | 1 | 0% | 2,265 | 1,713 | -24% | 0 | 0 | — |
case-13 | pass→pass | 14,764 | 9,580 | -35% | 1 | 1 | 0% | 2,410 | 1,910 | -21% | 0 | 0 | — |
case-14 | pass→pass | 14,364 | 6,733 | -53% | 1 | 1 | 0% | 2,181 | 1,389 | -36% | 0 | 0 | — |
case-15 | fail→pass | 7,183 | 7,955 | +11% | 1 | 1 | 0% | 1,129 | 1,565 | +39% | 0 | 0 | — |
case-16 | pass→pass | 11,107 | 7,836 | -29% | 1 | 1 | 0% | 1,870 | 1,437 | -23% | 0 | 0 | — |
case-17 | pass→pass | 5,844 | 4,702 | -20% | 1 | 1 | 0% | 798 | 1,010 | +27% | 0 | 0 | — |
case-18 | pass→pass | 12,700 | 8,168 | -36% | 1 | 1 | 0% | 2,044 | 1,470 | -28% | 0 | 0 | — |
case-19 | pass→pass | 8,869 | 8,978 | +1% | 1 | 1 | 0% | 1,353 | 1,858 | +37% | 0 | 0 | — |
case-20 | pass→pass | 10,412 | 11,054 | +6% | 1 | 1 | 0% | 1,491 | 2,127 | +43% | 0 | 0 | — |
case-21 | pass→pass | 4,925 | 9,817 | +99% | 1 | 1 | 0% | 727 | 1,272 | +75% | 0 | 0 | — |
case-22 | pass→pass | 10,032 | 9,268 | -8% | 1 | 1 | 0% | 1,577 | 1,712 | +9% | 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 +5 percentage points is the difference between those two pass rates over the 22 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.