Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Claude Code 세션을 위한 포괄적인 검증 시스템.
.claude/skills/loulanyue-verification-loop/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 1042% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-04 | ✓→✓ | = Same ✓ | -3% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 228% | 0% |
一个全面的 Claude Code 会话验证系统。
在以下情况下调用此技能:
bash# Check if project builds npm run build 2>&1 | tail -20 # OR pnpm build 2>&1 | tail -20
如果构建失败,请停止并在继续之前修复。
bash# TypeScript projects npx tsc --noEmit 2>&1 | head -30 # Python projects pyright . 2>&1 | head -30
报告所有类型错误。在继续之前修复关键错误。
bash# JavaScript/TypeScript npm run lint 2>&1 | head -30 # Python ruff check . 2>&1 | head -30
bash# Run tests with coverage npm run test -- --coverage 2>&1 | tail -50 # Check coverage threshold # Target: 80% minimum
报告:
bash# Check for secrets grep -rn "sk-" --include="*.ts" --include="*.js" . 2>/dev/null | head -10 grep -rn "api_key" --include="*.ts" --include="*.js" . 2>/dev/null | head -10 # Check for console.log grep -rn "console.log" --include="*.ts" --include="*.tsx" src/ 2>/dev/null | head -10
bash# Show what changed git diff --stat git diff HEAD~1 --name-only
审查每个更改的文件,检查:
运行所有阶段后,生成验证报告:
验证报告
==================
构建: [通过/失败]
类型: [通过/失败] (X 处错误)
代码检查: [通过/失败] (X 条警告)
测试: [通过/失败] (X/Y 通过,覆盖率 Z%)
安全: [通过/失败] (X 个问题)
差异: [X 个文件被修改]
总体: [就绪/未就绪] 提交 PR
待修复问题:
1. ...
2. ...对于长时间会话,每 15 分钟或在重大更改后运行验证:
markdown设置一个心理检查点: - 完成每个函数后 - 完成一个组件后 - 在移动到下一个任务之前 运行: /verify
此技能补充 PostToolUse 钩子,但提供更深入的验证。 钩子会立即捕获问题;此技能提供全面的审查。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 18,029 | 6,588 | -63% | 1 | 1 | 0% | 3,081 | 1,214 | -61% | 0 | 0 | — |
case-02 | fail→pass | 7,010 | 15,504 | +121% | 1 | 1 | 0% | 265 | 3,025 | +1042% | 0 | 0 | — |
case-03 | fail→fail | 16,582 | 6,131 | -63% | 1 | 1 | 0% | 1,308 | 1,042 | -20% | 0 | 0 | — |
case-04 | pass→pass | 10,539 | 4,690 | -55% | 1 | 1 | 0% | 1,641 | 1,595 | -3% | 0 | 0 | — |
case-09 | fail→pass | 8,623 | 4,560 | -47% | 1 | 1 | 0% | 1,517 | 1,215 | -20% | 0 | 0 | — |
case-05 | pass→pass | 2,103 | 2,358 | +12% | 1 | 1 | 0% | 348 | 1,141 | +228% | 0 | 0 | — |
case-06 | pass→pass | 4,093 | 2,578 | -37% | 1 | 1 | 0% | 561 | 1,121 | +100% | 0 | 0 | — |
case-07 | pass→pass | 4,170 | 2,533 | -39% | 1 | 1 | 0% | 711 | 1,148 | +61% | 0 | 0 | — |
case-08 | pass→pass | 9,474 | 2,151 | -77% | 1 | 1 | 0% | 1,619 | 1,153 | -29% | 0 | 0 | — |
case-10 | pass→pass | 4,023 | 2,851 | -29% | 1 | 1 | 0% | 655 | 1,213 | +85% | 0 | 0 | — |
case-11 | pass→pass | 7,446 | 4,188 | -44% | 1 | 1 | 0% | 1,335 | 1,442 | +8% | 0 | 0 | — |
case-12 | pass→pass | 3,821 | 1,921 | -50% | 1 | 1 | 0% | 564 | 1,083 | +92% | 0 | 0 | — |
case-13 | pass→pass | 15,232 | 12,713 | -17% | 1 | 1 | 0% | 2,208 | 2,723 | +23% | 0 | 0 | — |
case-14 | pass→pass | 2,836 | 2,069 | -27% | 1 | 1 | 0% | 409 | 1,092 | +167% | 0 | 0 | — |
case-15 | fail→pass | 8,824 | 4,148 | -53% | 1 | 1 | 0% | 1,374 | 1,565 | +14% | 0 | 0 | — |
case-16 | pass→pass | 12,379 | 5,359 | -57% | 1 | 1 | 0% | 1,913 | 1,696 | -11% | 0 | 0 | — |
case-17 | pass→pass | 11,787 | 3,319 | -72% | 1 | 1 | 0% | 1,962 | 1,341 | -32% | 0 | 0 | — |
case-18 | pass→pass | 6,796 | 2,423 | -64% | 1 | 1 | 0% | 1,004 | 1,160 | +16% | 0 | 0 | — |
case-19 | pass→pass | 17,469 | 16,168 | -7% | 1 | 1 | 0% | 2,610 | 3,272 | +25% | 0 | 0 | — |
case-20 | pass→pass | 5,053 | 5,482 | +8% | 1 | 1 | 0% | 1,003 | 1,896 | +89% | 0 | 0 | — |
case-21 | pass→pass | 5,697 | 4,307 | -24% | 1 | 1 | 0% | 979 | 1,573 | +61% | 0 | 0 | — |
case-22 | pass→pass | 13,645 | 14,449 | +6% | 1 | 1 | 0% | 2,704 | 3,916 | +45% | 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, and 19 counted toward the lift figure. The other 3 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 +14 percentage points is the difference between those two pass rates over the 19 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.