Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →你负责根据用户事实、capture、reference 和环境信息生成 `reference_contract`，并判定它是 `strict_ready`、`fallback_only` 还是 `missing`。
.claude/skills/haolange-skill-reference-contract/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -30% | 0% |
你负责根据用户事实、capture、reference 和环境信息生成 reference_contract，并判定它是 strict_ready、fallback_only 还是 missing。
reference_contract 是内部 contract，不是用户输入模板。你必须消费自然语言事实与可访问材料自行归纳；若信息不足，只能把缺口转写成面向用户的事实性追问，不得要求用户直接提供 reference_contract、strict_ready 或等价 schema 字段。
reference_contractreadiness_statusreadiness_rationalemissing_for_strict_ready| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 16,931 | 12,214 | -28% | 1 | 1 | 0% | 2,597 | 2,110 | -19% | 0 | 0 | — |
case-02 | fail→fail | 23,022 | 10,551 | -54% | 1 | 1 | 0% | 3,520 | 2,099 | -40% | 0 | 0 | — |
case-03 | fail→pass | 19,738 | 11,882 | -40% | 1 | 1 | 0% | 2,973 | 2,267 | -24% | 0 | 0 | — |
case-04 | fail→fail | 18,599 | 11,369 | -39% | 1 | 1 | 0% | 2,986 | 2,218 | -26% | 0 | 0 | — |
case-05 | fail→pass | 13,507 | 8,614 | -36% | 1 | 1 | 0% | 2,207 | 1,747 | -21% | 0 | 0 | — |
case-06 | fail→fail | 15,286 | 13,491 | -12% | 1 | 1 | 0% | 2,262 | 2,385 | +5% | 0 | 0 | — |
case-07 | fail→fail | 15,976 | 11,656 | -27% | 1 | 1 | 0% | 2,778 | 2,254 | -19% | 0 | 0 | — |
case-08 | fail→pass | 19,470 | 7,881 | -60% | 1 | 1 | 0% | 1,644 | 1,580 | -4% | 0 | 0 | — |
case-09 | fail→pass | 17,502 | 10,472 | -40% | 1 | 1 | 0% | 2,812 | 1,955 | -30% | 0 | 0 | — |
case-10 | fail→fail | 10,213 | 8,744 | -14% | 1 | 1 | 0% | 1,585 | 1,769 | +12% | 0 | 0 | — |
case-11 | fail→pass | 17,955 | 8,640 | -52% | 1 | 1 | 0% | 2,850 | 1,720 | -40% | 0 | 0 | — |
case-12 | fail→fail | 17,504 | 11,038 | -37% | 1 | 1 | 0% | 2,610 | 2,097 | -20% | 0 | 0 | — |
case-13 | fail→pass | 17,288 | 11,186 | -35% | 1 | 1 | 0% | 2,835 | 2,033 | -28% | 0 | 0 | — |
case-14 | fail→pass | 10,218 | 10,683 | +5% | 1 | 1 | 0% | 1,752 | 2,147 | +23% | 0 | 0 | — |
case-15 | fail→pass | 11,173 | 8,354 | -25% | 1 | 1 | 0% | 1,652 | 1,605 | -3% | 0 | 0 | — |
case-16 | fail→pass | 17,075 | 13,242 | -22% | 1 | 1 | 0% | 2,884 | 2,426 | -16% | 0 | 0 | — |
case-17 | fail→fail | 16,751 | 10,755 | -36% | 1 | 1 | 0% | 2,649 | 2,117 | -20% | 0 | 0 | — |
case-18 | fail→pass | 16,040 | 8,174 | -49% | 1 | 1 | 0% | 2,252 | 1,492 | -34% | 0 | 0 | — |
case-19 | fail→pass | 21,267 | 9,489 | -55% | 1 | 1 | 0% | 3,147 | 1,822 | -42% | 0 | 0 | — |
case-20 | pass→fail | 11,001 | 8,139 | -26% | 1 | 1 | 0% | 2,128 | 1,471 | -31% | 0 | 0 | — |
case-21 | pass→fail | 18,560 | 14,853 | -20% | 1 | 1 | 0% | 3,199 | 2,804 | -12% | 0 | 0 | — |
case-22 | pass→fail | 12,184 | 7,674 | -37% | 1 | 1 | 0% | 2,152 | 1,512 | -30% | 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 +41 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 cases got worse with the skill loaded, and they are 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.