Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →分析 GitHub Pull Request，评估必要性、描述完整性、验证证据、主要风险与是否可直接合入。
.claude/skills/charliedream1-analyze-pr/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -6% | 0% |
分析 GitHub Pull Request，评估必要性、描述完整性、验证证据、主要风险与是否可直接合入。
Repository: https://github.com/ZhuLinsen/daily_stock_analysis/pulls
text/analyze-pr <pr_number>
分析时使用简洁中文，优先遵循仓库根目录 AGENTS.md 和 .github/PULL_REQUEST_TEMPLATE.md。
bashgh pr view <pr_number> --repo ZhuLinsen/daily_stock_analysis gh pr view <pr_number> --repo ZhuLinsen/daily_stock_analysis --comments gh pr checks <pr_number> --repo ZhuLinsen/daily_stock_analysis gh pr diff <pr_number> --repo ZhuLinsen/daily_stock_analysis
如有失败的 CI，优先查看失败日志，而不是立刻在本地重跑全部检查：
bashgh run view <run_id> --log-failed
对照 .github/PULL_REQUEST_TEMPLATE.md，确认是否覆盖：
PR TypeBackground And ProblemScope Of ChangeIssue LinkVerification Commands And ResultsCompatibility And RiskRollback Plan若 PR 涉及第三方模型 / API 兼容语义、请求参数固定值、OpenAI-compatible 路由、YAML alias、fallback 行为或运行时配置保存 / 清理 / 迁移逻辑，还要额外检查描述里是否明确写出：
gh pr checks、PR diff、现有测试与工作流日志判断问题gh pr checkout如果必须补本地验证，按改动面选择最接近的检查，例如：
./scripts/ci_gate.sh 或 python -m py_compile <changed_python_files>cd apps/dsa-web && npm ci && npm run lint && npm run build重点检查：
保存到 .claude/reviews/prs/pr-<number>.md
markdown# PR #<number> Analysis **Date**: YYYY-MM-DD **Status**: Pending Review ## Findings - [严重级别] file:line - 问题描述 ## Summary - 必要性： - 是否有对应 issue： - PR 类型： - description 完整性： - 验证情况： - 主要风险： - 是否可直接合入： ## Validation Evidence - CI 结论： - 本地补充验证（如有）： ## Compatibility And Risk - API / Web / Desktop： - 配置 / Docker / GitHub Actions： - fallback / 通知 / 报告结构： - 第三方依赖 / 官方约束来源： - 运行时兼容窗口 / 已覆盖链路： - 旧配置迁移或静默改写风险： ## Draft Review Comment <建议评论内容>
执行以下动作前，先询问用户：
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 26,141 | 5,420 | -79% | 1 | 1 | 0% | 4,382 | 1,295 | -70% | 0 | 0 | — |
case-02 | fail→fail | 22,877 | 5,871 | -74% | 1 | 1 | 0% | 3,054 | 1,281 | -58% | 0 | 0 | — |
case-03 | fail→fail | 22,979 | 5,311 | -77% | 1 | 1 | 0% | 3,100 | 1,236 | -60% | 0 | 0 | — |
case-04 | fail→fail | 11,509 | 7,006 | -39% | 1 | 1 | 0% | 1,892 | 1,420 | -25% | 0 | 0 | — |
case-05 | fail→fail | 3,649 | 8,141 | +123% | 1 | 1 | 0% | 594 | 1,523 | +156% | 0 | 0 | — |
case-06 | fail→fail | 6,771 | 10,910 | +61% | 1 | 1 | 0% | 1,105 | 1,265 | +14% | 0 | 0 | — |
case-07 | pass→pass | 13,625 | 10,273 | -25% | 1 | 1 | 0% | 1,938 | 1,991 | +3% | 0 | 0 | — |
case-08 | pass→fail | 10,993 | 7,760 | -29% | 1 | 1 | 0% | 1,836 | 1,381 | -25% | 0 | 0 | — |
case-09 | fail→pass | 11,309 | 8,192 | -28% | 1 | 1 | 0% | 1,996 | 1,754 | -12% | 0 | 0 | — |
case-10 | pass→pass | 15,125 | 4,054 | -73% | 1 | 1 | 0% | 2,291 | 1,583 | -31% | 0 | 0 | — |
case-11 | fail→pass | 15,884 | 5,617 | -65% | 1 | 1 | 0% | 2,442 | 1,877 | -23% | 0 | 0 | — |
case-12 | fail→pass | 17,250 | 15,824 | -8% | 1 | 1 | 0% | 2,498 | 3,244 | +30% | 0 | 0 | — |
case-13 | fail→fail | 13,800 | 5,928 | -57% | 1 | 1 | 0% | 2,116 | 1,262 | -40% | 0 | 0 | — |
case-14 | fail→pass | 15,161 | 9,598 | -37% | 1 | 1 | 0% | 2,180 | 2,647 | +21% | 0 | 0 | — |
case-15 | pass→pass | 10,647 | 4,013 | -62% | 1 | 1 | 0% | 1,665 | 1,673 | +0% | 0 | 0 | — |
case-16 | fail→pass | 11,619 | 4,568 | -61% | 1 | 1 | 0% | 1,854 | 1,748 | -6% | 0 | 0 | — |
case-17 | fail→pass | 11,062 | 2,830 | -74% | 1 | 1 | 0% | 1,582 | 1,569 | -1% | 0 | 0 | — |
case-18 | fail→fail | 12,134 | 4,935 | -59% | 1 | 1 | 0% | 1,795 | 1,260 | -30% | 0 | 0 | — |
case-19 | fail→pass | 8,326 | 1,832 | -78% | 1 | 1 | 0% | 1,441 | 1,272 | -12% | 0 | 0 | — |
case-20 | fail→pass | 8,683 | 6,285 | -28% | 1 | 1 | 0% | 1,496 | 1,871 | +25% | 0 | 0 | — |
case-21 | fail→pass | 7,636 | 7,541 | -1% | 1 | 1 | 0% | 1,205 | 2,092 | +74% | 0 | 0 | — |
case-22 | fail→pass | 4,416 | 2,890 | -35% | 1 | 1 | 0% | 808 | 1,490 | +84% | 0 | 0 | — |
case-23 | fail→pass | 8,422 | 3,331 | -60% | 1 | 1 | 0% | 1,382 | 1,522 | +10% | 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 14 counted toward the lift figure. The other 9 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 +43 percentage points is the difference between those two pass rates over the 14 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.