Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股业绩超预期/低预期量化分析。当用户说"业绩超预期"、"earnings surprise"、"超预期"、"低预期"、"业绩打败预期"、"不及预期"时触发。量化分析业绩公告后的市场反应。支持formal和brief风格。
.claude/skills/aifinlab-a-share-earnings-surprise/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -27% | 0% |
bashSCRIPTS="$SKILLS_ROOT/cn-stock-data/scripts" python "$SCRIPTS/cn_stock_data.py" kline --code [CODE] --freq daily --start [日期] python "$SCRIPTS/cn_stock_data.py" quote --code [CODE] python "$SCRIPTS/cn_stock_data.py" finance --code [CODE]
分析业绩公告后的收益率漂移持续性
| 维度 | formal | brief | |------|--------|-------| | 超预期度 | SUE计算+排名 | 超/达/低预期 | | 市场反应 | CAR序列分析 | 公告后涨跌 | | 漂移分析 | PEAD统计 | 漂移方向 | 默认风格：brief。
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-06 | fail→pass | 13,474 | 3,242 | -76% | 1 | 1 | 0% | 1,883 | 1,093 | -42% | 0 | 0 | — |
case-01 | fail→fail | 27,135 | 12,155 | -55% | 1 | 1 | 0% | 4,264 | 1,128 | -74% | 0 | 0 | — |
case-02 | pass→fail | 20,905 | 39,837 | +91% | 1 | 1 | 0% | 3,279 | 1,081 | -67% | 0 | 0 | — |
case-03 | fail→fail | 33,082 | 8,984 | -73% | 1 | 1 | 0% | 4,366 | 1,034 | -76% | 0 | 0 | — |
case-04 | fail→pass | 13,048 | 5,392 | -59% | 1 | 1 | 0% | 2,342 | 1,478 | -37% | 0 | 0 | — |
case-05 | fail→pass | 13,873 | 7,396 | -47% | 1 | 1 | 0% | 1,978 | 1,285 | -35% | 0 | 0 | — |
case-07 | fail→fail | 23,674 | 15,488 | -35% | 1 | 1 | 0% | 3,288 | 2,652 | -19% | 0 | 0 | — |
case-08 | pass→fail | 18,625 | 9,200 | -51% | 1 | 1 | 0% | 1,927 | 868 | -55% | 0 | 0 | — |
case-09 | pass→pass | 23,579 | 21,081 | -11% | 1 | 1 | 0% | 3,216 | 3,429 | +7% | 0 | 0 | — |
case-10 | pass→pass | 5,791 | 7,086 | +22% | 1 | 1 | 0% | 994 | 1,426 | +43% | 0 | 0 | — |
case-11 | pass→pass | 18,805 | 21,687 | +15% | 1 | 1 | 0% | 3,058 | 3,569 | +17% | 0 | 0 | — |
case-12 | pass→pass | 19,418 | 15,244 | -21% | 1 | 1 | 0% | 2,669 | 2,741 | +3% | 0 | 0 | — |
case-13 | pass→pass | 13,515 | 14,349 | +6% | 1 | 1 | 0% | 1,794 | 2,563 | +43% | 0 | 0 | — |
case-14 | pass→pass | 13,186 | 11,942 | -9% | 1 | 1 | 0% | 1,856 | 2,198 | +18% | 0 | 0 | — |
case-15 | pass→fail | 15,389 | 7,834 | -49% | 1 | 1 | 0% | 2,461 | 843 | -66% | 0 | 0 | — |
case-16 | fail→pass | 18,217 | 24,940 | +37% | 1 | 1 | 0% | 2,826 | 4,510 | +60% | 0 | 0 | — |
case-17 | pass→pass | 8,345 | 3,520 | -58% | 1 | 1 | 0% | 1,343 | 1,123 | -16% | 0 | 0 | — |
case-18 | fail→pass | 9,724 | 2,901 | -70% | 1 | 1 | 0% | 1,221 | 890 | -27% | 0 | 0 | — |
case-19 | pass→pass | 31,171 | 16,288 | -48% | 1 | 1 | 0% | 4,527 | 3,389 | -25% | 0 | 0 | — |
case-20 | fail→fail | 38,892 | 7,892 | -80% | 1 | 1 | 0% | 6,361 | 862 | -86% | 0 | 0 | — |
case-21 | fail→fail | 20,792 | 9,852 | -53% | 1 | 1 | 0% | 2,932 | 983 | -66% | 0 | 0 | — |
case-22 | fail→fail | 21,050 | 8,002 | -62% | 1 | 1 | 0% | 3,339 | 846 | -75% | 0 | 0 | — |
case-23 | fail→fail | 32,282 | 40,909 | +27% | 1 | 1 | 0% | 6,207 | 8,714 | +40% | 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, and 15 counted toward the lift figure. The other 8 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +9 percentage points is the difference between those two pass rates over the 15 comparable cases. 4 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.