[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7979F7DP8H5MMT2CKE8R59D":3,"skill-same:01M7979F7DP8H5MMT2CKE8R59D":102},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":15,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":31,"commitSha":32,"skillDirectory":33,"forksCount":34,"repositoryHtmlUrl":35,"repositoryHomepageUrl":6,"ownerHtmlUrl":36,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":37,"install":38,"resources":51},"01M7979F7DP8H5MMT2CKE8R59D","skill-backtest-overfit",null,"Detect backtest overfitting and selection bias from multiple testing. Use when a user has a backtest \u002F factor result and asks whether the Sharpe is real, whether a strategy is overfit, or wants to validate results before trusting them. Computes Deflated Sharpe Ratio, Probability of Backtest Overfitting (PBO), multiple-testing haircut, and Minimum Track Record Length.","ahang1598\u002Fdoubao-workbuddy-qwenwork-skills","ahang1598","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F29569469?v=4","01M787BTDDPJN7G0C9TC9SQ85J","01M787BTE3YMSPS4EBYF2MCE1Z",98,"GPL-3.0","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\r\nname: skill-backtest-overfit\r\ndescription: \"Detect backtest overfitting and selection bias from multiple testing. Use when a user has a backtest \u002F factor result and asks whether the Sharpe is real, whether a strategy is overfit, or wants to validate results before trusting them. Computes Deflated Sharpe Ratio, Probability of Backtest Overfitting (PBO), multiple-testing haircut, and Minimum Track Record Length.\"\r\nlicense: GPL-3.0\r\ncategory: 工具\r\nmetadata:\r\n  organization: QuantSkills\r\n  organization_url: https:\u002F\u002Fgithub.com\u002Fquantskills\r\n  repository: skill-backtest-overfit\r\n  repository_url: https:\u002F\u002Fgithub.com\u002Fquantskills\u002Fskill-backtest-overfit\r\n  project_type: skill\r\n  collection: portfolio-risk-validation\r\n---\r\n\r\n## WorkBuddy 真实证据强制覆盖\r\n\r\n本节优先于本 Skill 后续示例。此成员只审计主流程封存的真实 `selected_returns.csv` 和\r\n`trials_matrix.csv`；不得自行连接本地数据 SDK，也不得用合成、模拟或随机行情替代缺失\r\n证据。若上游收益证据缺失，必须返回阻塞。需要补取市场数据时，按共享 Runtime Skill\r\n执行 `auth_status` 与 `call_pandadata`，保存真实结果后再进入统计审计。\r\n\r\n## WorkBuddy 真实证据强制覆盖\r\n\r\n本节优先于本 Skill 后续示例。此成员只审计主流程封存的真实 `selected_returns.csv` 和\r\n`trials_matrix.csv`；不得自行连接本地数据 SDK，也不得用合成、模拟或随机行情替代缺失\r\n证据。若上游收益证据缺失，必须返回阻塞。需要补取市场数据时，按共享 Runtime Skill\r\n执行 `auth_status` 与 `call_pandadata`，保存真实结果后再进入统计审计。\r\n\r\n## WorkBuddy 真实证据强制覆盖\r\n\r\n本节优先于本 Skill 后续示例。此成员只审计主流程封存的真实 `selected_returns.csv` 和\r\n`trials_matrix.csv`；不得自行连接本地数据 SDK，也不得用合成、模拟或随机行情替代缺失\r\n证据。若上游收益证据缺失，必须返回阻塞。需要补取市场数据时，按共享 Runtime Skill\r\n执行 `auth_status` 与 `call_pandadata`，保存真实结果后再进入统计审计。\r\n\r\n```json qsh-form\r\n{\r\n  \"version\": 1,\r\n  \"task\": {\r\n    \"placeholder\": \"说明待检验的回测、收益序列、试验次数及数据文件；可上传 selected_returns 和 trials_matrix\",\r\n    \"required\": true\r\n  },\r\n  \"fields\": [\r\n    {\r\n      \"key\": \"n_trials\",\r\n      \"label\": \"试验次数\",\r\n      \"type\": \"number\",\r\n      \"placeholder\": \"例如 200\",\r\n      \"help\": \"请如实填写所有尝试过的参数或策略配置数量\"\r\n    },\r\n    {\r\n      \"key\": \"periods_per_year\",\r\n      \"label\": \"年化期数\",\r\n      \"type\": \"number\",\r\n      \"default\": \"252\",\r\n      \"help\": \"日频通常为 252\"\r\n    },\r\n    {\r\n      \"key\": \"haircut_method\",\r\n      \"label\": \"多重检验方法\",\r\n      \"type\": \"select\",\r\n      \"default\": \"holm\",\r\n      \"options\": [\r\n        { \"value\": \"holm\", \"label\": \"Holm\" },\r\n        { \"value\": \"bonferroni\", \"label\": \"Bonferroni\" },\r\n        { \"value\": \"bhy\", \"label\": \"BHY\" }\r\n      ]\r\n    }\r\n  ],\r\n  \"prompt_template\": \"{{#task}}任务与材料：\\n{{task}}\\n\\n{{\u002Ftask}}{{#attachments}}用户上传的材料（已放入工作区）：\\n{{attachments}}\\n\\n{{\u002Fattachments}}检验该回测的过拟合与多重选择偏差。{{#n_trials}}按真实试验次数 {{n_trials}} 计算。{{\u002Fn_trials}}{{#periods_per_year}}年化期数采用 {{periods_per_year}}。{{\u002Fperiods_per_year}}{{#haircut_method}}多重检验校正采用 {{haircut_method}}。{{\u002Fhaircut_method}}计算 DSR、PBO、Haircut Sharpe 和最小样本长度，缺少试验矩阵时明确降级项及局限，给出可追溯的通过\u002F失败结论，输出中文报告。\"\r\n}\r\n```\r\n\r\n# skill-backtest-overfit\r\n\r\nrole: skill · output: OverfitReport (JSON + text) · paradigm: statistical validation of backtests\r\n\r\n把\"一个回测的 Sharpe 到底是不是真的\"变成可计算、可引用、可归档的统计结论。这是因子流水线最后、也是最缺的**统计守门员**。\r\n\r\n## 🎯 这个 Skill 解决什么问题\r\n\r\n量化生态在**批量造因子\u002F调参**：同一段历史数据上试了 N 个配置，挑出最好看的一个。问题是——**只要 N 够大，纯噪声里也必然能挑出一个 Sharpe 很高的\"策略\"**。常规 health-check（持仓数、换手率、回撤区间）抓不到这种选择偏差。\r\n\r\n本 Skill 用四个有原始文献支撑的统计量回答：\"这个回测有多大概率是过拟合？\"\r\n\r\n- **Deflated Sharpe Ratio (DSR)** — 把\"试了 N 次\"的选择偏差从 Sharpe 里扣掉，给出\"真 SR > 0\"的概率。\r\n- **PBO（回测过拟合概率）** — 样本内最优的配置，在样本外有多大概率沦为中等水平。\r\n- **Haircut Sharpe** — 多重检验下给 Sharpe 打的折。\r\n- **MinTRL** — 要多长样本，这个 Sharpe 才统计显著。\r\n\r\n## ⚡ 工作流（Agent 按此执行）\r\n\r\n1. **收集输入**：拿到①被选中策略的逐期收益序列；②（强烈建议）所有试过的配置组成的收益矩阵 `T×N`；③诚实的试验次数 `n_trials`（每个网格点、每个因子变体都算）。\r\n2. **算 DSR \u002F PSR \u002F MinTRL**：`scripts\u002Fdeflated_sharpe.py`。需要跨试验的 Sharpe 方差——有收益矩阵时自动估，没有则退化为单次估计（偏宽松，须提示用户）。\r\n3. **算 PBO**：`scripts\u002Fpbo_cscv.py`，需要 `T×N` 收益矩阵，CSCV 组合对称交叉验证。\r\n4. **算 Haircut**：`scripts\u002Fhaircut.py`，Bonferroni \u002F Holm \u002F BHY 三选一。\r\n5. **生成报告**：`scripts\u002Foverfit_report.py` 汇总四项 → `OverfitReport`（JSON + 文本），给出 PASS\u002FFAIL 与触发原因。\r\n6. **解释结论**：用业务语言告诉用户\"该回测约有 X% 概率是数据挖掘噪声\"，并指出最薄弱的统计项。\r\n\r\n```bash\r\n# 一行生成报告\r\npython scripts\u002Foverfit_report.py --returns selected.csv --trials trials.csv --n-trials 200 --out report.json\r\n# 离线演示（无需凭证，对比过拟合 vs 真实 edge）\r\n# WorkBuddy: use sealed real return files\r\n```\r\n\r\n## 🗃️ 输入契约\r\n\r\n| 输入 | 形态 | 必需 | 说明 |\r\n|------|------|------|------|\r\n| `selected_returns` | 1D 数组 \u002F CSV 单列 | 是 | 被选中策略的**逐期**收益（非年化）|\r\n| `trials_matrix` | `T×N` 矩阵 \u002F CSV | 否（PBO 必需）| 每列一个被试配置的逐期收益 |\r\n| `n_trials` | int | 是 | 诚实的多重检验次数 |\r\n| `periods_per_year` | int | 否 | 默认 252，用于年化展示 |\r\n\r\n输出 `OverfitReport`：`verdict \u002F passed \u002F deflated_sharpe_ratio \u002F pbo \u002F haircut \u002F minimum_track_record_length \u002F ...`\r\n\r\n## 🔗 管线定位\r\n\r\n```\r\n因子挖掘(批量) → 评估 → 回测(skill-backtest) → [本 Skill：过拟合统计守门] → 上线\r\n```\r\n\r\n它是 `skill-backtest` 的**统计补充**：`skill-backtest` 给净值与启发式 health-check，本 Skill 给多重检验下的统计显著性。**任一统计项不过，先怀疑是过拟合，不要相信回测。**\r\n\r\n## 📦 仓库结构\r\n\r\n```\r\nskill-backtest-overfit\u002F\r\n├── SKILL.md\r\n├── README.md\r\n├── requirements.txt\r\n├── scripts\u002F\r\n│   ├── deflated_sharpe.py   # DSR \u002F PSR \u002F MinTRL  (Bailey & Lopez de Prado 2014)\r\n│   ├── pbo_cscv.py          # PBO via CSCV        (Bailey et al. 2017)\r\n│   ├── purged_kfold.py      # 净化+禁运 K-Fold CV (Lopez de Prado 2018, ch.7)\r\n│   ├── haircut.py           # 多重检验 haircut    (Harvey & Liu 2015)\r\n│   ├── overfit_report.py    # 汇总 → OverfitReport\r\n├── references\u002F\r\n│   ├── methodology.md       # 公式推导 + 文献\r\n│   └── anti-patterns.md     # 10 种过拟合陷阱\r\n└── examples\u002F\r\n    └── run_demo.py          # 过拟合 vs 真实 edge 对照\r\n```\r\n\r\n## ⚠️ 使用规则\r\n\r\n- **诚实申报 `n_trials`** 是本 Skill 成立的前提：少报 = 自欺。\r\n- DSR\u002FPSR 用**逐期**Sharpe，勿传年化值。\r\n- PBO 需要足够多的配置列（N≥10 才有意义）；`n_blocks` 必须为偶数且 ≤ T。\r\n- 只做研究\u002F方法论参考，不构成投资建议。\r\n","## WorkBuddy 真实证据强制覆盖\r\n\r\n本节优先于本 Skill 后续示例。此成员只审计主流程封存的真实 `selected_returns.csv` 和\r\n`trials_matrix.csv`；不得自行连接本地数据 SDK，也不得用合成、模拟或随机行情替代缺失\r\n证据。若上游收益证据缺失，必须返回阻塞。需要补取市场数据时，按共享 Runtime Skill\r\n执行 `auth_status` 与 `call_pandadata`，保存真实结果后再进入统计审计。\r\n\r\n## WorkBuddy 真实证据强制覆盖\r\n\r\n本节优先于本 Skill 后续示例。此成员只审计主流程封存的真实 `selected_returns.csv` 和\r\n`trials_matrix.csv`；不得自行连接本地数据 SDK，也不得用合成、模拟或随机行情替代缺失\r\n证据。若上游收益证据缺失，必须返回阻塞。需要补取市场数据时，按共享 Runtime Skill\r\n执行 `auth_status` 与 `call_pandadata`，保存真实结果后再进入统计审计。\r\n\r\n## WorkBuddy 真实证据强制覆盖\r\n\r\n本节优先于本 Skill 后续示例。此成员只审计主流程封存的真实 `selected_returns.csv` 和\r\n`trials_matrix.csv`；不得自行连接本地数据 SDK，也不得用合成、模拟或随机行情替代缺失\r\n证据。若上游收益证据缺失，必须返回阻塞。需要补取市场数据时，按共享 Runtime Skill\r\n执行 `auth_status` 与 `call_pandadata`，保存真实结果后再进入统计审计。\r\n\r\n```json qsh-form\r\n{\r\n  \"version\": 1,\r\n  \"task\": {\r\n    \"placeholder\": \"说明待检验的回测、收益序列、试验次数及数据文件；可上传 selected_returns 和 trials_matrix\",\r\n    \"required\": true\r\n  },\r\n  \"fields\": [\r\n    {\r\n      \"key\": \"n_trials\",\r\n      \"label\": \"试验次数\",\r\n      \"type\": \"number\",\r\n      \"placeholder\": \"例如 200\",\r\n      \"help\": \"请如实填写所有尝试过的参数或策略配置数量\"\r\n    },\r\n    {\r\n      \"key\": \"periods_per_year\",\r\n      \"label\": \"年化期数\",\r\n      \"type\": \"number\",\r\n      \"default\": \"252\",\r\n      \"help\": \"日频通常为 252\"\r\n    },\r\n    {\r\n      \"key\": \"haircut_method\",\r\n      \"label\": \"多重检验方法\",\r\n      \"type\": \"select\",\r\n      \"default\": \"holm\",\r\n      \"options\": [\r\n        { \"value\": \"holm\", \"label\": \"Holm\" },\r\n        { \"value\": \"bonferroni\", \"label\": \"Bonferroni\" },\r\n        { \"value\": \"bhy\", \"label\": \"BHY\" }\r\n      ]\r\n    }\r\n  ],\r\n  \"prompt_template\": \"{{#task}}任务与材料：\\n{{task}}\\n\\n{{\u002Ftask}}{{#attachments}}用户上传的材料（已放入工作区）：\\n{{attachments}}\\n\\n{{\u002Fattachments}}检验该回测的过拟合与多重选择偏差。{{#n_trials}}按真实试验次数 {{n_trials}} 计算。{{\u002Fn_trials}}{{#periods_per_year}}年化期数采用 {{periods_per_year}}。{{\u002Fperiods_per_year}}{{#haircut_method}}多重检验校正采用 {{haircut_method}}。{{\u002Fhaircut_method}}计算 DSR、PBO、Haircut Sharpe 和最小样本长度，缺少试验矩阵时明确降级项及局限，给出可追溯的通过\u002F失败结论，输出中文报告。\"\r\n}\r\n```\r\n\r\n# skill-backtest-overfit\r\n\r\nrole: skill · output: OverfitReport (JSON + text) · paradigm: statistical validation of backtests\r\n\r\n把\"一个回测的 Sharpe 到底是不是真的\"变成可计算、可引用、可归档的统计结论。这是因子流水线最后、也是最缺的**统计守门员**。\r\n\r\n## 🎯 这个 Skill 解决什么问题\r\n\r\n量化生态在**批量造因子\u002F调参**：同一段历史数据上试了 N 个配置，挑出最好看的一个。问题是——**只要 N 够大，纯噪声里也必然能挑出一个 Sharpe 很高的\"策略\"**。常规 health-check（持仓数、换手率、回撤区间）抓不到这种选择偏差。\r\n\r\n本 Skill 用四个有原始文献支撑的统计量回答：\"这个回测有多大概率是过拟合？\"\r\n\r\n- **Deflated Sharpe Ratio (DSR)** — 把\"试了 N 次\"的选择偏差从 Sharpe 里扣掉，给出\"真 SR > 0\"的概率。\r\n- **PBO（回测过拟合概率）** — 样本内最优的配置，在样本外有多大概率沦为中等水平。\r\n- **Haircut Sharpe** — 多重检验下给 Sharpe 打的折。\r\n- **MinTRL** — 要多长样本，这个 Sharpe 才统计显著。\r\n\r\n## ⚡ 工作流（Agent 按此执行）\r\n\r\n1. **收集输入**：拿到①被选中策略的逐期收益序列；②（强烈建议）所有试过的配置组成的收益矩阵 `T×N`；③诚实的试验次数 `n_trials`（每个网格点、每个因子变体都算）。\r\n2. **算 DSR \u002F PSR \u002F MinTRL**：`scripts\u002Fdeflated_sharpe.py`。需要跨试验的 Sharpe 方差——有收益矩阵时自动估，没有则退化为单次估计（偏宽松，须提示用户）。\r\n3. **算 PBO**：`scripts\u002Fpbo_cscv.py`，需要 `T×N` 收益矩阵，CSCV 组合对称交叉验证。\r\n4. **算 Haircut**：`scripts\u002Fhaircut.py`，Bonferroni \u002F Holm \u002F BHY 三选一。\r\n5. **生成报告**：`scripts\u002Foverfit_report.py` 汇总四项 → `OverfitReport`（JSON + 文本），给出 PASS\u002FFAIL 与触发原因。\r\n6. **解释结论**：用业务语言告诉用户\"该回测约有 X% 概率是数据挖掘噪声\"，并指出最薄弱的统计项。\r\n\r\n```bash\r\n# 一行生成报告\r\npython scripts\u002Foverfit_report.py --returns selected.csv --trials trials.csv --n-trials 200 --out report.json\r\n# 离线演示（无需凭证，对比过拟合 vs 真实 edge）\r\n# WorkBuddy: use sealed real return files\r\n```\r\n\r\n## 🗃️ 输入契约\r\n\r\n| 输入 | 形态 | 必需 | 说明 |\r\n|------|------|------|------|\r\n| `selected_returns` | 1D 数组 \u002F CSV 单列 | 是 | 被选中策略的**逐期**收益（非年化）|\r\n| `trials_matrix` | `T×N` 矩阵 \u002F CSV | 否（PBO 必需）| 每列一个被试配置的逐期收益 |\r\n| `n_trials` | int | 是 | 诚实的多重检验次数 |\r\n| `periods_per_year` | int | 否 | 默认 252，用于年化展示 |\r\n\r\n输出 `OverfitReport`：`verdict \u002F passed \u002F deflated_sharpe_ratio \u002F pbo \u002F haircut \u002F minimum_track_record_length \u002F ...`\r\n\r\n## 🔗 管线定位\r\n\r\n```\r\n因子挖掘(批量) → 评估 → 回测(skill-backtest) → [本 Skill：过拟合统计守门] → 上线\r\n```\r\n\r\n它是 `skill-backtest` 的**统计补充**：`skill-backtest` 给净值与启发式 health-check，本 Skill 给多重检验下的统计显著性。**任一统计项不过，先怀疑是过拟合，不要相信回测。**\r\n\r\n## 📦 仓库结构\r\n\r\n```\r\nskill-backtest-overfit\u002F\r\n├── SKILL.md\r\n├── README.md\r\n├── requirements.txt\r\n├── scripts\u002F\r\n│   ├── deflated_sharpe.py   # DSR \u002F PSR \u002F MinTRL  (Bailey & Lopez de Prado 2014)\r\n│   ├── pbo_cscv.py          # PBO via CSCV        (Bailey et al. 2017)\r\n│   ├── purged_kfold.py      # 净化+禁运 K-Fold CV (Lopez de Prado 2018, ch.7)\r\n│   ├── haircut.py           # 多重检验 haircut    (Harvey & Liu 2015)\r\n│   ├── overfit_report.py    # 汇总 → OverfitReport\r\n├── references\u002F\r\n│   ├── methodology.md       # 公式推导 + 文献\r\n│   └── anti-patterns.md     # 10 种过拟合陷阱\r\n└── examples\u002F\r\n    └── run_demo.py          # 过拟合 vs 真实 edge 对照\r\n```\r\n\r\n## ⚠️ 使用规则\r\n\r\n- **诚实申报 `n_trials`** 是本 Skill 成立的前提：少报 = 自欺。\r\n- DSR\u002FPSR 用**逐期**Sharpe，勿传年化值。\r\n- PBO 需要足够多的配置列（N≥10 才有意义）；`n_blocks` 必须为偶数且 ≤ T。\r\n- 只做研究\u002F方法论参考，不构成投资建议。\r\n",{"name":5,"license":14,"category":24,"metadata":25,"description":7},"工具",{"collection":26,"repository":5,"organization":27,"project_type":28,"repository_url":29,"organization_url":30},"portfolio-risk-validation","QuantSkills","skill","https:\u002F\u002Fgithub.com\u002Fquantskills\u002Fskill-backtest-overfit","https:\u002F\u002Fgithub.com\u002Fquantskills","main","5e45653f59a2e4a2a65f575e8aaf7637165674d5","workbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit",39,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills","https:\u002F\u002Fgithub.com\u002Fahang1598","2026-11-06 15:01:27",{"source":8,"skillName":5,"commands":39,"prompt":49,"zipAvailable":50},[40,43,46],{"tool":41,"command":42},"npx","npx skills add ahang1598\u002Fdoubao-workbuddy-qwenwork-skills --skill skill-backtest-overfit",{"tool":44,"command":45},"bunx","bunx skills add ahang1598\u002Fdoubao-workbuddy-qwenwork-skills --skill skill-backtest-overfit",{"tool":47,"command":48},"pnpm","pnpm dlx skills add ahang1598\u002Fdoubao-workbuddy-qwenwork-skills --skill skill-backtest-overfit","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002FSKILL.md",true,[52,58,62,68,72,76,80,86,90,94,98],{"path":53,"resourceType":54,"resourceTypeDesc":55,"sizeBytes":56,"sourceHtmlUrl":57},".gitignore","OTHER","其他",148,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002F.gitignore",{"path":59,"resourceType":59,"resourceTypeDesc":60,"sizeBytes":61,"sourceHtmlUrl":15},"LICENSE","许可证",35823,{"path":63,"resourceType":64,"resourceTypeDesc":65,"sizeBytes":66,"sourceHtmlUrl":67},"README.md","REFERENCE","参考文档",198,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002FREADME.md",{"path":69,"resourceType":64,"resourceTypeDesc":65,"sizeBytes":70,"sourceHtmlUrl":71},"references\u002Fanti-patterns.md",1561,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002Freferences\u002Fanti-patterns.md",{"path":73,"resourceType":64,"resourceTypeDesc":65,"sizeBytes":74,"sourceHtmlUrl":75},"references\u002Fmethodology.md",4393,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002Freferences\u002Fmethodology.md",{"path":77,"resourceType":64,"resourceTypeDesc":65,"sizeBytes":78,"sourceHtmlUrl":79},"requirements.txt",124,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002Frequirements.txt",{"path":81,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":84,"sourceHtmlUrl":85},"scripts\u002Fdeflated_sharpe.py","SCRIPT","脚本",7920,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002Fscripts\u002Fdeflated_sharpe.py",{"path":87,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":88,"sourceHtmlUrl":89},"scripts\u002Fhaircut.py",4406,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002Fscripts\u002Fhaircut.py",{"path":91,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":92,"sourceHtmlUrl":93},"scripts\u002Foverfit_report.py",6838,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002Fscripts\u002Foverfit_report.py",{"path":95,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":96,"sourceHtmlUrl":97},"scripts\u002Fpbo_cscv.py",5570,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002Fscripts\u002Fpbo_cscv.py",{"path":99,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":100,"sourceHtmlUrl":101},"scripts\u002Fpurged_kfold.py",4420,"https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fexperts\u002Fpandaai-ai-quant-research-team\u002Fskills\u002Fskill-backtest-overfit\u002Fscripts\u002Fpurged_kfold.py",[103,112,121,128,136,144],{"skillNo":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":107,"licenseFileUrl":108,"sourceHtmlUrl":109,"categories":110,"keywords":111,"publishedAt":19},"01M79AD9QRFFSMWE9ZHAV6049W","xiaoyi-pdf-win","PDF 综合处理技能，处理文档生成、编辑、安全与解析。 适用情形： 1. 创建与排版：从零生成报告、提案、简历等 PDF，或对现有文档重新排版美化； 2. 表单与水印：自动填写 PDF 表单字段，或添加文字\u002F图片水印（如打水印、标机密）； 3. 页面管理：合并多个 PDF，或拆分、提取指定页码； 4. 安全控制：为 PDF 添加密码（加密）或移除密码（解密）； 5. 内容提取：从 PDF 中提取纯文本或导出表格数据。 只要用户诉求涉及生成、排版、美化、转换、拼接、拆分 PDF，或处理水印、表单、密码，必须触发本技能。","MIT","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fxiaoyiwork\u002Fskills\u002Fxiaoyi-pdf-win\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fxiaoyiwork\u002Fskills\u002Fxiaoyi-pdf-win\u002FSKILL.md",[],[],{"skillNo":113,"name":114,"nameZh":6,"description":115,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":116,"licenseFileUrl":117,"sourceHtmlUrl":118,"categories":119,"keywords":120,"publishedAt":19},"01M79A6AXW44V3H0DVPHXE32YE","skill-creator-win","Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.","Apache-2.0","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fxiaoyiwork\u002Fskills\u002Fskill-creator-win\u002Flicense.txt","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fxiaoyiwork\u002Fskills\u002Fskill-creator-win\u002FSKILL.md",[],[],{"skillNo":122,"name":123,"nameZh":6,"description":124,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":107,"licenseFileUrl":6,"sourceHtmlUrl":125,"categories":126,"keywords":127,"publishedAt":19},"01M79A5DRYQNS61T7W9TA1ZNPY","markitdown-win","Convert files and office documents to Markdown. Supports PDF, DOCX, PPTX, XLSX, images (with OCR), audio (with transcription), HTML, CSV, JSON, XML, ZIP, YouTube URLs, EPubs and more.","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fxiaoyiwork\u002Fskills\u002Fmarkitdown-win\u002FSKILL.md",[],[],{"skillNo":129,"name":130,"nameZh":6,"description":131,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":107,"licenseFileUrl":132,"sourceHtmlUrl":133,"categories":134,"keywords":135,"publishedAt":19},"01M79A35XSZGH60W5XQE6CR0Q5","minimax-docx","Professional DOCX document creation, editing, and formatting using OpenXML SDK (.NET). Three pipelines: (A) create new documents from scratch, (B) fill\u002Fedit content in existing documents, (C) apply template formatting with XSD validation gate-check. MUST use this skill whenever the user wants to produce, modify, or format a Word document — including when they say \"write a report\", \"draft a proposal\", \"make a contract\", \"fill in this form\", \"reformat to match this template\", or any task whose final output is a .docx file. Even if the user doesn't mention \"docx\" explicitly, if the task implies a printable\u002Fformal document, use this skill.","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fxiaoyiwork\u002Fexperts\u002Fdocument-generation-expert\u002Fskills\u002Fminimax-docx\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fxiaoyiwork\u002Fexperts\u002Fdocument-generation-expert\u002Fskills\u002Fminimax-docx\u002FSKILL.md",[],[],{"skillNo":137,"name":138,"nameZh":6,"description":139,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":107,"licenseFileUrl":140,"sourceHtmlUrl":141,"categories":142,"keywords":143,"publishedAt":19},"01M79A1WEZ6HQWMH6KDE32Q928","skillhub-daily","SkillHub 每日推荐 - 扫描 skillhub.cn 全站 Top100 + 7 大分类各 Top20（共 240 个 Skill），\n基于用户痛点精准推荐高价值 Skill，支持 IMA\u002F飞书\u002FObsidian 多通道存储。\n触发词：每日推荐、SkillHub 日报、潜力Skill、帮我推荐技能。\n支持两种使用模式：常规 Skills 模式（对话触发）\u002F Cron 定时任务模式。","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fskills\u002Fskillhub-daily\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fskills\u002Fskillhub-daily\u002FSKILL.md",[],[],{"skillNo":145,"name":146,"nameZh":6,"description":147,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":107,"licenseFileUrl":6,"sourceHtmlUrl":148,"categories":149,"keywords":150,"publishedAt":19},"01M799ZCAYFF5A73KRC4G4VMWX","deep-research","Structured deep research workflow with human-in-the-loop control. Use \u002Fresearch to generate research outline, \u002Fresearch-deep for parallel web search across items, \u002Fresearch-report to compile markdown reports. Supports academic research, benchmark research, technology selection, market analysis, and due diligence. Triggers: 'deep research', 'research topic', 'benchmark comparison', 'technology survey', 'market analysis'. Requires: WebSearch capability.","https:\u002F\u002Fgithub.com\u002Fahang1598\u002Fdoubao-workbuddy-qwenwork-skills\u002Fblob\u002Fmain\u002Fworkbuddy\u002Fskills\u002Fdeep-research\u002FSKILL.md",[],[]]