Install any skill in seconds. Free to start, no credit card required.
Get Started Free →经济责任审计中的资产管理审计。审查国有资产保值增值、资产处置合规性、 对外投资收益、产权变动等。
.claude/skills/thomasmoreai-asset-review/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 26% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -37% | 0% |
你是资产审计师,负责审查领导干部任期内国有资产的管理、使用和处置情况, 重点关注国有资产保值增值。
bashls .auditstack/scope/audit-plan.md 2>/dev/null && echo "审计方案已就绪" || echo "建议先运行 /scope" ls .auditstack/financial/findings.md 2>/dev/null && echo "财务审计底稿已就绪" || echo "建议先运行 /financial-review"
核心指标:
| 指标 | 计算方式 | 关注点 | |------|----------|--------| | 国有资产保值增值率 | 期末净资产 ÷ 期初净资产 × 100% | <100% 需查明原因 | | 资产负债率 | 负债总额 ÷ 资产总额 × 100% | 异常升高需关注 | | 净资产收益率 | 净利润 ÷ 平均净资产 × 100% | 低于同行业水平 |
审查要点:
审查要点:
重点关注:
低价转让 → 是否存在利益输送?
无偿划转 → 审批程序是否合规?
资产报废 → 是否提前报废可用资产?
资产置换 → 换入换出价值是否对等?审查要点:
审查要点:
审查要点:
markdown## 资产审计发现 [编号]:[标题] **资产类型:** [固定资产/无形资产/对外投资/在建工程/其他] **审计事项:** [具体审查内容] **事实描述:** [客观描述,含金额、时间、涉及资产明细] **审计依据:** [违反的规定] **对国有资产的影响:** [保值增值/流失/潜在风险] **涉及金额:** [资产原值、评估价、处置价、损失金额] **责任归属:** [直接责任/主管责任/领导责任] **审计证据:** [产权证、评估报告、审批文件、合同等]
bashmkdir -p .auditstack/asset
保存到 .auditstack/asset/ 目录:
findings.md — 资产审计发现asset-summary.md — 国有资产保值增值情况汇总表| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 22,164 | 21,849 | -1% | 1 | 1 | 0% | 3,821 | 5,144 | +35% | 0 | 0 | — |
case-02 | fail→pass | 24,434 | 24,671 | +1% | 1 | 1 | 0% | 3,832 | 5,114 | +33% | 0 | 0 | — |
case-03 | fail→pass | 27,342 | 24,692 | -10% | 1 | 1 | 0% | 4,213 | 5,294 | +26% | 0 | 0 | — |
case-04 | pass→pass | 21,474 | 29,895 | +39% | 1 | 1 | 0% | 3,656 | 4,488 | +23% | 0 | 0 | — |
case-05 | pass→pass | 21,985 | 27,565 | +25% | 1 | 1 | 0% | 3,415 | 5,423 | +59% | 0 | 0 | — |
case-06 | pass→fail | 21,774 | 23,679 | +9% | 1 | 1 | 0% | 3,331 | 4,817 | +45% | 0 | 0 | — |
case-07 | pass→pass | 16,136 | 17,880 | +11% | 1 | 1 | 0% | 2,535 | 3,805 | +50% | 0 | 0 | — |
case-08 | pass→pass | 16,014 | 15,125 | -6% | 1 | 1 | 0% | 2,541 | 3,550 | +40% | 0 | 0 | — |
case-09 | fail→pass | 20,087 | 15,017 | -25% | 1 | 1 | 0% | 3,129 | 3,589 | +15% | 0 | 0 | — |
case-10 | fail→pass | 21,147 | 16,843 | -20% | 1 | 1 | 0% | 3,785 | 3,803 | +0% | 0 | 0 | — |
case-11 | pass→pass | 21,591 | 19,913 | -8% | 1 | 1 | 0% | 3,209 | 4,002 | +25% | 0 | 0 | — |
case-12 | pass→pass | 17,485 | 16,454 | -6% | 1 | 1 | 0% | 2,630 | 3,531 | +34% | 0 | 0 | — |
case-13 | pass→pass | 19,465 | 18,780 | -4% | 1 | 1 | 0% | 2,892 | 3,781 | +31% | 0 | 0 | — |
case-14 | fail→fail | 19,877 | 19,756 | -1% | 1 | 1 | 0% | 3,165 | 4,274 | +35% | 0 | 0 | — |
case-15 | pass→pass | 15,230 | 15,136 | -1% | 1 | 1 | 0% | 2,315 | 3,401 | +47% | 0 | 0 | — |
case-16 | fail→fail | 23,070 | 21,568 | -7% | 1 | 1 | 0% | 3,340 | 4,329 | +30% | 0 | 0 | — |
case-17 | fail→pass | 17,308 | 5,040 | -71% | 1 | 1 | 0% | 2,859 | 1,792 | -37% | 0 | 0 | — |
case-18 | pass→pass | 17,727 | 17,069 | -4% | 1 | 1 | 0% | 2,663 | 3,508 | +32% | 0 | 0 | — |
case-19 | pass→pass | 20,079 | 19,360 | -4% | 1 | 1 | 0% | 3,338 | 4,202 | +26% | 0 | 0 | — |
case-20 | pass→pass | 22,138 | 22,150 | +0% | 1 | 1 | 0% | 2,722 | 4,498 | +65% | 0 | 0 | — |
case-21 | fail→pass | 17,967 | 3,813 | -79% | 1 | 1 | 0% | 3,020 | 1,638 | -46% | 0 | 0 | — |
case-22 | pass→pass | 21,427 | 15,331 | -28% | 1 | 1 | 0% | 3,116 | 3,211 | +3% | 0 | 0 | — |
case-23 | pass→pass | 17,394 | 18,393 | +6% | 1 | 1 | 0% | 2,426 | 3,495 | +44% | 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 +22 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.