Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Stripeなどの接続された請求ツールを使用して、サブスクリプション、返金、チャーントリアージ、請求ポータルの回復、プラン分析などの顧客請求ワークフローを操作します。顧客を助けたい、サブスクリプション状態を検査したい、または収益に影響する請求操作を管理したい場合に使用します。
.claude/skills/affaan-m-customer-billing-ops/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 4% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 83% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -6% | 0% |
此技能用于真实的客户运营操作，而非通用的支付 API 设计。
目标是帮助运营人员回答：客户是谁、发生了什么、最安全的修复方案是什么、以及后续应发送什么跟进内容。
从最可靠的标识符入手：
返回简洁的身份摘要：
在操作前将案例归入一个类别：
| 案例 | 典型操作 | |------|----------------| | 重复的个人订阅 | 取消多余订阅，考虑退款 | | 真实的多席位/团队意图 | 保留席位，澄清计费模式 | | 支付失败/结账不完整 | 通过门户恢复或更新支付方式 | | 缺少自助控制功能 | 提供门户、取消路径或发票访问权限 | | 产品故障或信任破裂 | 退款、道歉、记录产品问题 |
推荐顺序：
若修复需要产品工作，需区分：
若客户痛点源于缺少运营界面，需明确指出。常见示例：
将这些视为 ECC 或网站跟进事项，而非单纯的支持事件。
最终需包含：
使用以下结构：
text客户 - 姓名 / 邮箱 - 相关账户标识 计费状态 - 活跃订阅 - 发票或续费状态 - 异常情况 决策 - 问题分类 - 为何此操作正确 已执行操作 - 退款 / 取消 / 门户 / 无操作 后续跟进 - 简短客户消息 产品缺口 - 产品或网站中应修复的内容
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 16,012 | 10,636 | -34% | 1 | 1 | 0% | 2,539 | 2,645 | +4% | 0 | 0 | — |
case-02 | fail→pass | 17,151 | 12,394 | -28% | 1 | 1 | 0% | 2,759 | 2,933 | +6% | 0 | 0 | — |
case-03 | pass→fail | 15,905 | 13,340 | -16% | 1 | 1 | 0% | 3,107 | 3,342 | +8% | 0 | 0 | — |
case-04 | pass→pass | 14,421 | 13,101 | -9% | 1 | 1 | 0% | 2,881 | 3,315 | +15% | 0 | 0 | — |
case-05 | pass→pass | 14,666 | 15,093 | +3% | 1 | 1 | 0% | 3,269 | 4,224 | +29% | 0 | 0 | — |
case-06 | pass→fail | 19,781 | 15,747 | -20% | 1 | 1 | 0% | 3,010 | 3,206 | +7% | 0 | 0 | — |
case-07 | pass→pass | 18,262 | 10,798 | -41% | 1 | 1 | 0% | 2,654 | 2,696 | +2% | 0 | 0 | — |
case-08 | pass→pass | 15,967 | 13,687 | -14% | 1 | 1 | 0% | 2,522 | 3,000 | +19% | 0 | 0 | — |
case-09 | fail→pass | 11,128 | 14,743 | +32% | 1 | 1 | 0% | 1,792 | 3,286 | +83% | 0 | 0 | — |
case-10 | fail→pass | 17,629 | 11,082 | -37% | 1 | 1 | 0% | 2,614 | 2,734 | +5% | 0 | 0 | — |
case-11 | pass→pass | 17,818 | 10,593 | -41% | 1 | 1 | 0% | 2,750 | 2,704 | -2% | 0 | 0 | — |
case-12 | fail→pass | 18,728 | 11,240 | -40% | 1 | 1 | 0% | 2,806 | 2,648 | -6% | 0 | 0 | — |
case-13 | fail→pass | 18,972 | 12,297 | -35% | 1 | 1 | 0% | 2,924 | 3,000 | +3% | 0 | 0 | — |
case-14 | fail→fail | 18,603 | 12,937 | -30% | 1 | 1 | 0% | 2,598 | 3,036 | +17% | 0 | 0 | — |
case-15 | fail→fail | 20,788 | 12,903 | -38% | 1 | 1 | 0% | 3,078 | 2,885 | -6% | 0 | 0 | — |
case-16 | fail→fail | 17,480 | 9,302 | -47% | 1 | 1 | 0% | 2,677 | 2,569 | -4% | 0 | 0 | — |
case-17 | pass→pass | 16,008 | 12,456 | -22% | 1 | 1 | 0% | 2,327 | 2,752 | +18% | 0 | 0 | — |
case-18 | fail→pass | 16,659 | 13,254 | -20% | 1 | 1 | 0% | 2,538 | 3,035 | +20% | 0 | 0 | — |
case-19 | pass→pass | 20,544 | 16,599 | -19% | 1 | 1 | 0% | 3,056 | 2,444 | -20% | 0 | 0 | — |
case-20 | fail→pass | 14,556 | 11,452 | -21% | 1 | 1 | 0% | 2,128 | 2,653 | +25% | 0 | 0 | — |
case-21 | fail→pass | 18,534 | 16,926 | -9% | 1 | 1 | 0% | 2,750 | 2,674 | -3% | 0 | 0 | — |
case-22 | fail→pass | 13,531 | 10,997 | -19% | 1 | 1 | 0% | 1,994 | 2,605 | +31% | 0 | 0 | — |
case-23 | pass→fail | 13,285 | 12,286 | -8% | 1 | 1 | 0% | 2,048 | 2,835 | +38% | 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. 23 cases were attempted. The headline lift of +30 percentage points is the difference between those two pass rates over the 23 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.