Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →需求转译 - 将主观、模糊、口语化的用户需求翻译为结构化、精确的AI可执行规格。 This skill should be used when the user gives a vague, subjective, or ambiguous requirement, or when the user explicitly requests requirement translation. Triggers: 需求转译, 翻译需求, 明确需求, 细化需求, "帮我分析一下这个需求", "这个需求不够清楚"
.claude/skills/aiskillstore-requirement-translator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 26% | 0% |
将主观模糊的用户需求翻译为结构化、精确的 AI 可执行规格。
以下任一情况触发本 Skill：
从原始需求中提取：
将需求转译为以下结构化格式。若某章节无内容，注明「无」而非省略。
## 需求规格书
### 1. 目标概述
[一句话描述核心目标，点明要解决的问题和预期结果]
### 2. 功能需求
- F1: [具体可执行的功能描述]
- F2: [具体可执行的功能描述]
### 3. 技术约束
- T1: [技术栈 / 平台 / 兼容性要求]
- T2: [性能 / 安全 / 规模约束]
### 4. 设计约束
- D1: [视觉风格 / 交互模式 / 品牌约束]
- D2: [布局 / 响应式 / 无障碍要求]
### 5. 验收标准
- AC1: [可客观验证的完成标准，含具体数值或状态]
- AC2: [可客观验证的完成标准]
### 6. 边界条件与异常处理
- E1: [异常场景及预期处理方式]
- E2: [极端输入 / 空状态 / 错误状态处理]
### 7. 优先级
- P0 必须: [核心不可妥协的需求]
- P1 应该: [重要但可后续迭代]
- P2 可选: [锦上添花，可省略]在转译结果后附加：
[推断][推断] 标记，方便用户审查和纠正| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 24,241 | 17,526 | -28% | 1 | 1 | 0% | 3,086 | 2,758 | -11% | 0 | 0 | — |
case-01 | fail→pass | 31,268 | 29,497 | -6% | 1 | 1 | 0% | 4,101 | 4,615 | +13% | 0 | 0 | — |
case-02 | fail→pass | 26,543 | 28,948 | +9% | 1 | 1 | 0% | 4,175 | 4,584 | +10% | 0 | 0 | — |
case-03 | fail→pass | 31,039 | 29,891 | -4% | 1 | 1 | 0% | 5,199 | 5,015 | -4% | 0 | 0 | — |
case-05 | pass→fail | 24,360 | 22,246 | -9% | 1 | 1 | 0% | 4,023 | 3,685 | -8% | 0 | 0 | — |
case-06 | pass→fail | 22,735 | 31,840 | +40% | 1 | 1 | 0% | 3,493 | 5,281 | +51% | 0 | 0 | — |
case-07 | fail→fail | 24,685 | 18,071 | -27% | 1 | 1 | 0% | 4,158 | 3,590 | -14% | 0 | 0 | — |
case-08 | fail→pass | 21,114 | 18,495 | -12% | 1 | 1 | 0% | 2,320 | 3,690 | +59% | 0 | 0 | — |
case-09 | fail→pass | 25,831 | 24,731 | -4% | 1 | 1 | 0% | 3,239 | 4,078 | +26% | 0 | 0 | — |
case-10 | fail→pass | 24,128 | 16,826 | -30% | 1 | 1 | 0% | 3,281 | 3,569 | +9% | 0 | 0 | — |
case-11 | fail→fail | 43,610 | 30,275 | -31% | 1 | 1 | 0% | 4,103 | 4,176 | +2% | 0 | 0 | — |
case-12 | pass→pass | 28,192 | 31,741 | +13% | 1 | 1 | 0% | 4,608 | 3,889 | -16% | 0 | 0 | — |
case-13 | fail→fail | 24,365 | 37,193 | +53% | 1 | 1 | 0% | 3,314 | 4,210 | +27% | 0 | 0 | — |
case-14 | fail→pass | 25,247 | 21,188 | -16% | 1 | 1 | 0% | 3,220 | 3,347 | +4% | 0 | 0 | — |
case-15 | fail→pass | 20,696 | 17,485 | -16% | 1 | 1 | 0% | 3,411 | 3,631 | +6% | 0 | 0 | — |
case-16 | pass→pass | 39,978 | 39,388 | -1% | 1 | 1 | 0% | 2,537 | 3,862 | +52% | 0 | 0 | — |
case-17 | pass→pass | 36,620 | 30,048 | -18% | 1 | 1 | 0% | 3,203 | 4,591 | +43% | 0 | 0 | — |
case-18 | fail→pass | 38,736 | 27,507 | -29% | 1 | 1 | 0% | 3,895 | 4,321 | +11% | 0 | 0 | — |
case-19 | pass→pass | 25,723 | 34,888 | +36% | 1 | 1 | 0% | 3,468 | 4,182 | +21% | 0 | 0 | — |
case-20 | fail→fail | 34,309 | 27,613 | -20% | 1 | 1 | 0% | 2,866 | 4,062 | +42% | 0 | 0 | — |
case-21 | pass→pass | 24,582 | 43,164 | +76% | 1 | 1 | 0% | 2,778 | 4,012 | +44% | 0 | 0 | — |
case-22 | pass→pass | 42,896 | 43,530 | +1% | 1 | 1 | 0% | 3,994 | 4,879 | +22% | 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 +32 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 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.