Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →监管问询回复助手（债券版）- 协助撰写债券审核问询回复。当用户需要为公司债、企业债等债券项目准备交易所/交易商协会问询回复、梳理回复思路或组织回复材料时触发此技能。适用于投行、律所、会所等中介机构的债券承销场景。
.claude/skills/thomasmoreai-regulatory-inquiry-response-bond/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-23 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-08 | ✓→✓ | ＝ Same ✓ | 51% | 0% |
| case-10 | ✓→✓ | ＝ Same ✓ | 53% | 0% |
使用此技能当：
首先向用户确认以下信息：
将问询问题按以下类型分类：
为每个问题生成回复框架：
markdown## 问题 [X]：[问题标题] ### 问题原文 [审核机构问询原文] ### 回复要点 1. [核心观点 1] 2. [核心观点 2] 3. [核心观点 3] ### 回复正文 #### 一、[分论点 1] [详细论述 + 数据支撑] #### 二、[分论点 2] [详细论述 + 数据支撑] ### 核查程序 1. [核查程序 1] 2. [核查程序 2] ### 核查结论 [明确结论] ### 附件材料 - [附件 1 名称] - [附件 2 名称]
城投类发行人：
房地产发行人：
| 问题类型 | 回复策略 | |----------|----------| | 资产负债率偏高 | 说明行业特点、展示现金流覆盖、融资空间 | | 盈利能力下滑 | 分析原因、展示改善措施、行业对比 | | 短期债务集中到期 | 展示偿债安排、授信储备、再融资计划 | | 对外担保较多 | 披露被担保方资质、反担保措施、代偿风险 | | 募集资金用途不明 | 细化资金使用计划、说明必要性、披露管理制度 |
markdown## 问题 [X] 核查材料清单 ### 财务核查 - [ ] 审计报告 - [ ] 财务报表明细 - [ ] 银行流水 - [ ] 征信报告 ### 业务核查 - [ ] 主要合同 - [ ] 客户/供应商访谈 - [ ] 行业数据 ### 合规核查 - [ ] 征信查询 - [ ] 诉讼查询 - [ ] 行政处罚查询 - [ ] 合规证明
references/bond-inquiry-types.md - 债券问询问题类型库references/response-templates-bond.md - 债券回复模板references/bond-rules.md - 债券发行规则汇编references/credit-analysis.md - 信用分析方法| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 24,687 | 22,769 | -8% | 1 | 1 | 0% | 3,778 | 4,826 | +28% | 0 | 0 | — |
case-02 | fail→fail | 25,211 | 27,877 | +11% | 1 | 1 | 0% | 3,622 | 5,399 | +49% | 0 | 0 | — |
case-03 | fail→fail | 22,733 | 28,461 | +25% | 1 | 1 | 0% | 3,532 | 5,917 | +68% | 0 | 0 | — |
case-04 | fail→fail | 28,848 | 24,239 | -16% | 1 | 1 | 0% | 4,337 | 5,505 | +27% | 0 | 0 | — |
case-05 | fail→fail | 30,324 | 24,669 | -19% | 1 | 1 | 0% | 5,066 | 5,434 | +7% | 0 | 0 | — |
case-06 | fail→fail | 24,232 | 21,541 | -11% | 1 | 1 | 0% | 3,517 | 4,781 | +36% | 0 | 0 | — |
case-07 | fail→fail | 22,194 | 18,542 | -16% | 1 | 1 | 0% | 3,205 | 4,313 | +35% | 0 | 0 | — |
case-08 | pass→pass | 22,176 | 22,269 | +0% | 1 | 1 | 0% | 3,094 | 4,686 | +51% | 0 | 0 | — |
case-09 | fail→pass | 21,918 | 22,089 | +1% | 1 | 1 | 0% | 3,231 | 4,636 | +43% | 0 | 0 | — |
case-10 | pass→pass | 17,770 | 16,893 | -5% | 1 | 1 | 0% | 2,676 | 4,106 | +53% | 0 | 0 | — |
case-11 | pass→pass | 15,769 | 17,379 | +10% | 1 | 1 | 0% | 2,406 | 3,993 | +66% | 0 | 0 | — |
case-12 | pass→pass | 18,309 | 17,184 | -6% | 1 | 1 | 0% | 2,887 | 4,061 | +41% | 0 | 0 | — |
case-13 | fail→pass | 18,187 | 20,247 | +11% | 1 | 1 | 0% | 2,804 | 4,248 | +51% | 0 | 0 | — |
case-14 | pass→pass | 14,919 | 15,985 | +7% | 1 | 1 | 0% | 2,287 | 3,965 | +73% | 0 | 0 | — |
case-15 | pass→pass | 15,512 | 12,614 | -19% | 1 | 1 | 0% | 2,609 | 3,460 | +33% | 0 | 0 | — |
case-16 | pass→pass | 9,731 | 6,713 | -31% | 1 | 1 | 0% | 1,478 | 2,415 | +63% | 0 | 0 | — |
case-17 | pass→pass | 14,352 | 13,375 | -7% | 1 | 1 | 0% | 2,238 | 3,541 | +58% | 0 | 0 | — |
case-18 | pass→pass | 9,057 | 3,268 | -64% | 1 | 1 | 0% | 1,371 | 1,943 | +42% | 0 | 0 | — |
case-19 | pass→pass | 7,936 | 7,049 | -11% | 1 | 1 | 0% | 1,228 | 2,417 | +97% | 0 | 0 | — |
case-20 | pass→pass | 9,969 | 5,627 | -44% | 1 | 1 | 0% | 1,644 | 2,273 | +38% | 0 | 0 | — |
case-21 | pass→pass | 12,892 | 10,944 | -15% | 1 | 1 | 0% | 1,875 | 2,989 | +59% | 0 | 0 | — |
case-22 | pass→pass | 18,812 | 17,248 | -8% | 1 | 1 | 0% | 2,763 | 4,005 | +45% | 0 | 0 | — |
case-23 | fail→pass | 12,510 | 4,645 | -63% | 1 | 1 | 0% | 2,001 | 2,147 | +7% | 0 | 0 | — |
case-24 | pass→pass | 16,571 | 13,711 | -17% | 1 | 1 | 0% | 2,607 | 3,584 | +37% | 0 | 0 | — |
case-25 | pass→pass | 19,420 | 18,925 | -3% | 1 | 1 | 0% | 2,867 | 4,213 | +47% | 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. 25 cases were attempted. The headline lift of +12 percentage points is the difference between those two pass rates over the 25 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.