Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →当需要审核营销或客服话术的合规性、识别禁用表述并输出修订建议时使用本技能，适用于合规审核与质检场景。
.claude/skills/aifinlab-bank-t226-compliance-operations-talk-track-review-assistant/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 42% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 241% | 0% |
本技能用于对营销/客服话术进行合规审核，识别违规表述、缺失风险提示或夸大收益等问题，并输出修订建议与升级路径。
scripts/talk_track_review_checker.py| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→fail | 31,077 | 27,977 | -10% | 1 | 1 | 0% | 4,394 | 4,417 | +1% | 0 | 0 | — |
case-01 | pass→pass | 15,327 | 16,326 | +7% | 1 | 1 | 0% | 2,519 | 3,240 | +29% | 0 | 0 | — |
case-02 | fail→pass | 15,588 | 13,429 | -14% | 1 | 1 | 0% | 2,006 | 2,572 | +28% | 0 | 0 | — |
case-03 | pass→pass | 14,789 | 17,488 | +18% | 1 | 1 | 0% | 2,192 | 3,061 | +40% | 0 | 0 | — |
case-04 | fail→pass | 11,057 | 14,244 | +29% | 1 | 1 | 0% | 1,666 | 2,361 | +42% | 0 | 0 | — |
case-06 | pass→pass | 14,675 | 17,562 | +20% | 1 | 1 | 0% | 2,056 | 2,847 | +38% | 0 | 0 | — |
case-07 | fail→pass | 14,365 | 21,350 | +49% | 1 | 1 | 0% | 2,229 | 2,965 | +33% | 0 | 0 | — |
case-08 | fail→pass | 16,961 | 14,215 | -16% | 1 | 1 | 0% | 2,088 | 2,530 | +21% | 0 | 0 | — |
case-09 | pass→pass | 14,358 | 15,322 | +7% | 1 | 1 | 0% | 2,162 | 2,481 | +15% | 0 | 0 | — |
case-10 | fail→pass | 6,435 | 18,281 | +184% | 1 | 1 | 0% | 847 | 2,886 | +241% | 0 | 0 | — |
case-11 | fail→pass | 5,094 | 15,851 | +211% | 1 | 1 | 0% | 696 | 2,666 | +283% | 0 | 0 | — |
case-12 | pass→pass | 12,575 | 14,852 | +18% | 1 | 1 | 0% | 1,605 | 2,390 | +49% | 0 | 0 | — |
case-13 | pass→pass | 20,354 | 20,369 | +0% | 1 | 1 | 0% | 3,231 | 3,198 | -1% | 0 | 0 | — |
case-14 | pass→pass | 16,294 | 15,806 | -3% | 1 | 1 | 0% | 2,167 | 2,445 | +13% | 0 | 0 | — |
case-15 | pass→pass | 12,805 | 15,608 | +22% | 1 | 1 | 0% | 1,908 | 2,502 | +31% | 0 | 0 | — |
case-16 | fail→fail | 13,955 | 15,637 | +12% | 1 | 1 | 0% | 1,865 | 2,602 | +40% | 0 | 0 | — |
case-17 | fail→pass | 17,517 | 17,886 | +2% | 1 | 1 | 0% | 2,187 | 2,720 | +24% | 0 | 0 | — |
case-18 | pass→pass | 12,324 | 14,047 | +14% | 1 | 1 | 0% | 1,882 | 2,262 | +20% | 0 | 0 | — |
case-19 | fail→fail | 15,077 | 14,998 | -1% | 1 | 1 | 0% | 2,442 | 2,775 | +14% | 0 | 0 | — |
case-20 | fail→pass | 13,481 | 10,035 | -26% | 1 | 1 | 0% | 1,946 | 1,861 | -4% | 0 | 0 | — |
case-21 | pass→pass | 15,106 | 12,646 | -16% | 1 | 1 | 0% | 2,205 | 2,452 | +11% | 0 | 0 | — |
case-22 | fail→fail | 14,075 | 13,397 | -5% | 1 | 1 | 0% | 1,783 | 2,205 | +24% | 0 | 0 | — |
case-23 | fail→pass | 18,514 | 19,833 | +7% | 1 | 1 | 0% | 2,759 | 3,016 | +9% | 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 +35 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is 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.