本仓库包含 11 项可安装的公共核心技能包。每项都提供可执行的 SKILL.md 主工作流、agents/openai.yaml Agent 接口、方法参考 references/、明确标注的合成训练案例和可复用模板 templates/。
git clone https://github.com/aierlanjiu/xuemu-creative-skills.git cd xuemu-creative-skills # 查看可用技能 sh install.sh --list # 安装单个技能至 ~/.codex/skills sh install.sh automotive-content-master # 安装全量技能 (自动时间戳备份) sh install.sh --all --backup-existing
点击任意行查看工作流摘要与调用方式；通过文件入口进入完整主流程、参考、示例和模板
Use $automotive-content-master to turn this EV thermal runaway protection topic into a structured engineering investigation and content package.
Use $automotive-content-master-lite to turn "SiC Inverter Efficiency" into a concise article package with 5 visual prompts and sources.
Use $build-powertrain-nvh-content to investigate 48th-order e-drive whine issue using source-path-receiver analysis and evidence grading.
Use $guizang-social-card-skill to format this workflow guide into a 5-page Xiaohongshu 3:4 carousel card set.
Use $lead-magnet-packager-agent to package these raw notes into an aligned, high-value PDF handbook.
Use $polishing-text-for-readability to rewrite this complex paragraph into clear, short, rhythmic Chinese sentences.
Use $powertrain-patent-content-publisher to interpret published DHT transmission patent CN11xxxxxxxA into an analysis package.
Use $qiaomu-xinzhiyuan-title to generate 5 high-density, fact-anchored title candidates for this AI chip article.
Use $rebuilding-selfmedia-monetization to audit this creator business funnel and design a 30-day monetization plan.
Use $wechat-infocard to format this battery tech brief into a WeChat inline-styled HTML layout.
Use $xuemu-illustration-core to turn "AI should narrow choices" into a 3:4 hand-drawn explainer prompt and shot plan.
按任务顺序组合技能，逐阶段验收研究、内容、视觉与交付物
适用于复杂汽车工程选题。从物理机制推导、五图视觉规划，到微信公众号内联 HTML 与事实锚定标题。
适用于创作者变现规划。开展 RAMP 诊断与报价控制，再把 Markdown 笔记整理为承诺覆盖矩阵、可编辑源文件与可选 PDF。
适用于动力总成专利解读。完成公开状态判定与权利要求提炼，并生成暖纸手绘风格的生图提示词。
遵循渐进披露与标准 Agent 接口规范
skills/automotive-content-master/
├── SKILL.md # 事实来源与渐进披露主流程
├── agents/
│ └── openai.yaml # OpenAI/Codex 界面元数据与调用策略
├── references/ # 工作表与判定框架 (按需读取)
│ ├── engineering-judgment-framework.md
│ └── visual-evidence-mapping.md
├── examples/ # 合成训练案例 (Synthetic)
│ └── synthetic-ev-thermal-investigation.md
└── templates/ # 高价值证据登记表模板
└── evidence-register.md
agents/openai.yaml 在支持该字段的 OpenAI/Codex 环境中提供界面元数据与调用策略；其他 Agent 产品应以 SKILL.md 为主：
interface: display_name: "Automotive Content Master" short_description: "Build source-backed automotive content packages." default_prompt: "Use $automotive-content-master to turn..." policy: allow_implicit_invocation: true
已脱敏的独立开源代码仓库与在线专题教程 Pages
明确划分公共仓库与本地/私有环境的安全隔离边界
| 安全边界维度 | 公共仓库包含 (Public Included) | 非公开/本地版隔离 (Private / Local Only) |
|---|---|---|
| 密钥与凭据 | 零密钥、零 Token、零 Cookie | 本地开发环境变量与 API Key |
| 案例数据 | 明确标注的合成训练案例 | 真实客户商业项目、合同与交付件 |
| 路径与配置 | 相对路径与标准环境变量 ($CODEX_HOME) | 硬编码的本机绝对路径与内部服务器 IP |
| 专利与交底 | 已公开专利的解读方法、结构示意计划与引用规则 | 未公开的技术交底草案与保密图纸 |