Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Translate PRD intent, roadmap asks, or product discussions into an implementation-ready capability plan that exposes constraints, invariants, interfaces, and unresolved decisions before multi-service work starts. Use when the user needs an ECC-native PRD-to-SRS lane instead of vague planning prose.
.claude/skills/affaan-m-product-capability/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 17% | 0% |
该技能将产品意图转化为明确的工程约束。
当问题不在于"我们应该构建什么？"，而在于"在开始实现之前，必须明确哪些条件？"时使用。
如果仓库中存在持久化的产品上下文文件，例如 PRODUCT.md、docs/product/ 或程序规范目录，请在此处更新。
如果尚不存在能力清单，请使用以下模板创建：
docs/examples/product-capability-template.md目标不是创建另一个规划栈，而是使隐藏的能力约束变得持久且可复用。
仅读取必要内容：
PRODUCT.md、设计文档、RFC、迁移笔记、运营模式文档将需求压缩为一个精确的陈述：
如果此陈述薄弱，实现将会偏离方向。
提取实现前必须满足的约束：
这些往往是仅存在于高级工程师记忆中的内容。
制定一份SRS风格的能力计划，包含：
以精确的交接点结束：
如有帮助，可指向下一个ECC原生通道：
project-flow-opsworkspace-surface-auditapi-connector-builderdashboard-buildertdd-workflowverification-loop按以下顺序返回结果：
text能力 - 一段重新陈述 约束条件 - 固定规则、不变项和边界 实现契约 - 参与者 - 界面 - 状态与转换 - 接口/数据影响 非目标 - 该通道明确不负责的内容 待定问题 - 仍需解决的阻碍或产品决策 交接 - 下一步应执行的操作及应由哪个ECC通道负责
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | fail→pass | 19,336 | 17,649 | -9% | 1 | 1 | 0% | 3,133 | 3,693 | +18% | 0 | 0 | — |
case-01 | fail→pass | 23,553 | 19,932 | -15% | 1 | 1 | 0% | 3,856 | 4,389 | +14% | 0 | 0 | — |
case-02 | fail→pass | 24,619 | 19,337 | -21% | 1 | 1 | 0% | 4,205 | 4,435 | +5% | 0 | 0 | — |
case-03 | fail→pass | 25,026 | 18,996 | -24% | 1 | 1 | 0% | 4,579 | 4,314 | -6% | 0 | 0 | — |
case-04 | fail→pass | 19,989 | 17,872 | -11% | 1 | 1 | 0% | 3,475 | 4,072 | +17% | 0 | 0 | — |
case-05 | fail→pass | 16,068 | 17,744 | +10% | 1 | 1 | 0% | 2,606 | 3,830 | +47% | 0 | 0 | — |
case-06 | fail→pass | 20,108 | 19,509 | -3% | 1 | 1 | 0% | 3,679 | 4,201 | +14% | 0 | 0 | — |
case-07 | fail→fail | 17,423 | 15,871 | -9% | 1 | 1 | 0% | 3,007 | 3,535 | +18% | 0 | 0 | — |
case-08 | fail→pass | 23,734 | 18,269 | -23% | 1 | 1 | 0% | 3,685 | 4,079 | +11% | 0 | 0 | — |
case-09 | pass→pass | 20,276 | 19,036 | -6% | 1 | 1 | 0% | 3,553 | 3,879 | +9% | 0 | 0 | — |
case-10 | fail→pass | 19,452 | 17,904 | -8% | 1 | 1 | 0% | 3,370 | 4,045 | +20% | 0 | 0 | — |
case-11 | pass→pass | 20,942 | 18,087 | -14% | 1 | 1 | 0% | 3,469 | 3,967 | +14% | 0 | 0 | — |
case-12 | fail→pass | 15,034 | 14,644 | -3% | 1 | 1 | 0% | 2,735 | 3,530 | +29% | 0 | 0 | — |
case-13 | pass→pass | 22,281 | 18,086 | -19% | 1 | 1 | 0% | 3,484 | 3,736 | +7% | 0 | 0 | — |
case-14 | fail→pass | 22,284 | 17,330 | -22% | 1 | 1 | 0% | 3,831 | 3,760 | -2% | 0 | 0 | — |
case-15 | fail→pass | 19,678 | 20,805 | +6% | 1 | 1 | 0% | 3,614 | 4,507 | +25% | 0 | 0 | — |
case-16 | fail→pass | 17,197 | 15,181 | -12% | 1 | 1 | 0% | 2,947 | 3,479 | +18% | 0 | 0 | — |
case-18 | fail→fail | 15,762 | 13,919 | -12% | 1 | 1 | 0% | 2,892 | 3,506 | +21% | 0 | 0 | — |
case-19 | pass→pass | 18,204 | 16,255 | -11% | 1 | 1 | 0% | 3,153 | 3,999 | +27% | 0 | 0 | — |
case-20 | pass→fail | 18,159 | 8,410 | -54% | 1 | 1 | 0% | 3,497 | 2,337 | -33% | 0 | 0 | — |
case-21 | pass→pass | 8,188 | 5,145 | -37% | 1 | 1 | 0% | 1,547 | 1,808 | +17% | 0 | 0 | — |
case-22 | pass→fail | 15,752 | 13,074 | -17% | 1 | 1 | 0% | 3,369 | 3,525 | +5% | 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 +50 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.