Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →异常交易检测助手。专注于识别账户异常交易行为，包括频繁交易、大额交易、关联交易、内幕交易风险等，帮助合规风控人员及时发现潜在风险。 **触发场景**： - 用户需要检测账户异常交易行为 - 用户说"异常交易"、"交易检测"、"风控筛查" - 需要识别频繁交易、大额交易、关联交易 - 需要内幕交易风险筛查、合规检查 **关键词**："异常交易"、"交易检测"、"风控"、"合规"、"频繁交易"、"大额交易"、"关联交易"、"内幕交易"
.claude/skills/aifinlab-abnormal-trading-detection-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 23% | 0% |
你是一名经验丰富的合规风控专家，擅长识别账户异常交易行为，帮助机构及时发现潜在合规风险。
| 类型 | 定义 | 检测指标 | |------|------|----------| | 频繁交易 | 短时间内大量交易 | 日交易次数、换手率 | | 大额交易 | 单笔/累计金额过大 | 单笔金额、累计金额 | | 关联交易 | 与关联方交易 | 关联关系、交易价格 | | 内幕交易 | 敏感期交易 | 时间窗口、信息敏感度 | | 操纵市场 | 异常价格/成交量 | 价格偏离、成交量异常 |
markdown# 【异常交易检测报告】 ## 检测账户 - 账户名称：[名称] - 检测期间：[起止日期] - 检测类型：[类型] ## 异常交易识别 ### [异常类型 1] - 触发规则：[规则描述] - 异常交易：[交易明细] - 风险等级：[高/中/低] - 建议措施：[措施] ### [异常类型 2] - [同上] ## 风险汇总 | 异常类型 | 次数 | 风险等级 | 状态 | |----------|------|----------|------| | | | | | ## 建议措施 - [措施 1] - [措施 2] ## 后续跟踪 - [ ] [跟踪事项 1] - [ ] [跟踪事项 2] --- 检测人：[姓名] 检测时间：[时间]
| 等级 | 标准 | 措施 | |------|------|------| | 高 | 涉嫌违法违规 | 立即报告、暂停交易 | | 中 | 异常明显、需关注 | 加强监控、询问客户 | | 低 | 轻微异常、可解释 | 记录备案、持续观察 |
输出前自查：
python# 调用 skill result = run_skill({ "param1": "value1", "param2": "value2" })
bashpython scripts/run_skill.py --input data.json
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 29,540 | 29,664 | +0% | 1 | 1 | 0% | 4,615 | 4,383 | -5% | 0 | 0 | — |
case-02 | fail→pass | 21,652 | 15,262 | -30% | 1 | 1 | 0% | 3,557 | 3,554 | -0% | 0 | 0 | — |
case-03 | fail→fail | 26,484 | 21,968 | -17% | 1 | 1 | 0% | 4,204 | 4,521 | +8% | 0 | 0 | — |
case-04 | pass→pass | 24,421 | 26,859 | +10% | 1 | 1 | 0% | 3,384 | 4,550 | +34% | 0 | 0 | — |
case-05 | pass→pass | 26,059 | 21,391 | -18% | 1 | 1 | 0% | 3,833 | 4,329 | +13% | 0 | 0 | — |
case-06 | pass→pass | 24,550 | 27,565 | +12% | 1 | 1 | 0% | 3,762 | 5,181 | +38% | 0 | 0 | — |
case-07 | pass→pass | 25,066 | 27,208 | +9% | 1 | 1 | 0% | 3,742 | 5,412 | +45% | 0 | 0 | — |
case-08 | pass→pass | 26,434 | 19,411 | -27% | 1 | 1 | 0% | 3,766 | 3,831 | +2% | 0 | 0 | — |
case-09 | fail→pass | 23,798 | 20,082 | -16% | 1 | 1 | 0% | 3,148 | 4,086 | +30% | 0 | 0 | — |
case-10 | pass→pass | 27,033 | 18,347 | -32% | 1 | 1 | 0% | 3,437 | 3,913 | +14% | 0 | 0 | — |
case-11 | pass→pass | 27,475 | 15,164 | -45% | 1 | 1 | 0% | 3,329 | 3,401 | +2% | 0 | 0 | — |
case-12 | pass→pass | 26,052 | 21,049 | -19% | 1 | 1 | 0% | 3,487 | 4,163 | +19% | 0 | 0 | — |
case-13 | pass→pass | 11,674 | 14,170 | +21% | 1 | 1 | 0% | 1,878 | 2,818 | +50% | 0 | 0 | — |
case-14 | pass→pass | 19,422 | 16,887 | -13% | 1 | 1 | 0% | 2,813 | 3,468 | +23% | 0 | 0 | — |
case-15 | pass→pass | 22,559 | 16,101 | -29% | 1 | 1 | 0% | 3,349 | 3,397 | +1% | 0 | 0 | — |
case-16 | fail→fail | 25,677 | 18,971 | -26% | 1 | 1 | 0% | 3,341 | 3,411 | +2% | 0 | 0 | — |
case-17 | fail→fail | 21,427 | 16,324 | -24% | 1 | 1 | 0% | 3,301 | 3,591 | +9% | 0 | 0 | — |
case-18 | pass→pass | 19,580 | 16,827 | -14% | 1 | 1 | 0% | 2,908 | 3,299 | +13% | 0 | 0 | — |
case-19 | fail→pass | 27,129 | 22,326 | -18% | 1 | 1 | 0% | 3,792 | 4,292 | +13% | 0 | 0 | — |
case-20 | pass→pass | 25,679 | 23,666 | -8% | 1 | 1 | 0% | 4,039 | 4,157 | +3% | 0 | 0 | — |
case-21 | fail→pass | 21,473 | 18,874 | -12% | 1 | 1 | 0% | 3,331 | 4,087 | +23% | 0 | 0 | — |
case-22 | fail→pass | 17,697 | 16,311 | -8% | 1 | 1 | 0% | 2,938 | 3,802 | +29% | 0 | 0 | — |
case-23 | pass→pass | 14,707 | 18,809 | +28% | 1 | 1 | 0% | 2,222 | 3,967 | +79% | 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 +26 percentage points is the difference between those two pass rates over the 23 comparable cases.
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.