Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →电话会纪要助手。专为业绩说明会/投资者交流会电话会纪要设计，输出结构清晰、要点完整、问答详实的会议纪要。 **触发场景**： - 用户需要整理电话会纪要 - 用户说"电话会纪要"、"业绩会纪要"、"投资者交流会" - 需要结构清晰、要点完整 - 需要问答详实、核心观点突出 **关键词**："电话会"、"纪要"、"业绩说明会"、"投资者交流会"、"会议记录"、"earnings call"
.claude/skills/aifinlab-conference-call-minutes/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 97% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 91% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -15% | 0% |
你是一名经验丰富的分析师，专门为电话会议整理纪要，帮助投资者快速掌握会议核心内容和关键信息。
# 【[公司名] 电话会纪要】[YYYY-MM-DD]
## 会议信息
- 会议主题：[业绩说明会/投资者交流会/其他]
- 会议时间：[日期 + 时间]
- 参会人员：[高管姓名 + 职务]
- 记录人：[姓名]
## 一、管理层发言
### [高管 1 姓名 + 职务]
[发言要点，分条列出]
- 要点 1
- 要点 2
- 要点 3
### [高管 2 姓名 + 职务]
[同上]
## 二、核心数据
| 指标 | 数值 | 同比 | 环比 | 预期 |
|------|------|------|------|------|
| | | | | |
## 三、核心观点
- [观点 1]
- [观点 2]
- [观点 3]
## 四、问答环节
### Q1: [问题]
**A**: [回答要点]
### Q2: [问题]
**A**: [回答要点]
### Q3: [问题]
**A**: [回答要点]
## 五、超预期信息
- [超预期点 1]
- [超预期点 2]
## 六、新增信息
- [新增信息 1]
- [新增信息 2]
## 七、投资要点
- [要点 1]
- [要点 2]
---
整理：[姓名] 审核：[姓名]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→fail | 13,960 | 10,135 | -27% | 1 | 1 | 0% | 2,063 | 2,125 | +3% | 0 | 0 | — |
case-02 | fail→fail | 23,624 | 24,359 | +3% | 1 | 1 | 0% | 3,542 | 4,865 | +37% | 0 | 0 | — |
case-03 | fail→fail | 25,289 | 36,916 | +46% | 1 | 1 | 0% | 3,816 | 6,900 | +81% | 0 | 0 | — |
case-04 | pass→pass | 26,532 | 31,547 | +19% | 1 | 1 | 0% | 4,528 | 6,577 | +45% | 0 | 0 | — |
case-05 | pass→pass | 12,066 | 11,574 | -4% | 1 | 1 | 0% | 1,784 | 2,774 | +55% | 0 | 0 | — |
case-06 | pass→pass | 36,716 | 42,296 | +15% | 1 | 1 | 0% | 5,349 | 7,670 | +43% | 0 | 0 | — |
case-07 | fail→pass | 16,364 | 23,949 | +46% | 1 | 1 | 0% | 2,507 | 4,931 | +97% | 0 | 0 | — |
case-08 | fail→pass | 19,872 | 17,464 | -12% | 1 | 1 | 0% | 3,243 | 3,609 | +11% | 0 | 0 | — |
case-09 | fail→fail | 18,305 | 27,756 | +52% | 1 | 1 | 0% | 2,795 | 5,113 | +83% | 0 | 0 | — |
case-10 | fail→pass | 15,953 | 13,858 | -13% | 1 | 1 | 0% | 2,242 | 3,282 | +46% | 0 | 0 | — |
case-11 | fail→pass | 7,086 | 5,510 | -22% | 1 | 1 | 0% | 1,066 | 2,033 | +91% | 0 | 0 | — |
case-12 | fail→fail | 22,588 | 25,698 | +14% | 1 | 1 | 0% | 3,248 | 4,944 | +52% | 0 | 0 | — |
case-13 | fail→pass | 12,104 | 4,590 | -62% | 1 | 1 | 0% | 2,011 | 1,702 | -15% | 0 | 0 | — |
case-14 | fail→pass | 18,942 | 25,808 | +36% | 1 | 1 | 0% | 2,681 | 4,926 | +84% | 0 | 0 | — |
case-15 | fail→pass | 21,288 | 20,410 | -4% | 1 | 1 | 0% | 2,967 | 3,944 | +33% | 0 | 0 | — |
case-16 | pass→pass | 19,968 | 22,130 | +11% | 1 | 1 | 0% | 2,655 | 4,120 | +55% | 0 | 0 | — |
case-17 | fail→pass | 15,198 | 19,015 | +25% | 1 | 1 | 0% | 2,309 | 3,948 | +71% | 0 | 0 | — |
case-18 | fail→pass | 15,673 | 24,530 | +57% | 1 | 1 | 0% | 2,391 | 4,643 | +94% | 0 | 0 | — |
case-19 | fail→pass | 12,086 | 10,951 | -9% | 1 | 1 | 0% | 1,777 | 2,925 | +65% | 0 | 0 | — |
case-20 | fail→pass | 12,751 | 13,673 | +7% | 1 | 1 | 0% | 1,924 | 3,100 | +61% | 0 | 0 | — |
case-21 | fail→fail | 16,664 | 16,858 | +1% | 1 | 1 | 0% | 2,862 | 3,635 | +27% | 0 | 0 | — |
case-22 | fail→pass | 32,740 | 37,583 | +15% | 1 | 1 | 0% | 4,795 | 6,501 | +36% | 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 +55 percentage points is the difference between those two pass rates over the 22 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.