Meta-skill

Execute skill chain plans generated by skill-chain-planner (v2.0). Reads the chain plan directory, parses the two-layer execution model, reliability design, degradation matrix, and execution state machine, then recursively calls `skill-for-skills` to generate each sub-skill. Orchestrates creation order by priority and dependency, with state-machine tracking, budget guard, degradation handling, and reliability conformance validation. Triggered by: "执行skill链", "创建skill链", "执行链规划", "批量创建skill", "chain executor", "执行规划", "递归生成skill", "按照规划创建", "deploy chain", "execute plan", "create chain skills", "可靠性验证", "降级处理", "状态机续接", "预算守卫".

Automatically generate, upgrade, or self-update standards for Claude Code skills based on user requirements. Reads sum.md for writing standards, searches the web to verify external APIs and tools, and creates or updates production-ready SKILL.md and README.md files with proper frontmatter, workflow instructions, and examples. Can also search official docs to update its own skill writing standards (sum.md). Triggered by: "编写skill", "生成skill", "创建skill", "写skill", "自动生成skill", "skill生成", "create skill", "generate skill", "升级skill", "更新skill", "修改skill", "添加功能", "upgrade skill", "更新规范", "更新sum.md", "同步官方文档", "update standards", "帮我写一个skill", "帮我生成一个skill", "skill-for-skills".

Decompose complex tasks into multi-skill chains with a two-layer execution model (deterministic orchestrator + inner tool-call loop), reliability pillars (cache-first / tool-call repair / cost & budget guard), feedback loops, degradation matrix, execution state machine, and security review. Generates step-by-step plans for using `skill-for-skills` to construct each skill and compose them into a working pipeline. Triggered by: "任务分解", "skill链规划", "复杂任务拆分", "多skill协作", "chain planner", "工作流拆分", "多步骤任务", "skill pipeline", "架构规划", "多步工作流", "pipeline设计", "skill依赖分析", "skill编排", "任务流水线", "可靠性规划", "降级矩阵", "执行状态机", "反馈循环", "预算守卫", "两层执行模型".

Build and maintain a local vulnerability database by looking up CVE information from the web and organizing it into a structured year/CVE hierarchy. Searches NVD, MITRE, GitHub Security Advisories, Exploit-DB, CISA KEV, and more to compile structured reports with standardized templates. Maintains a searchable database index with metadata for each CVE entry. Triggered by: "CVE", "cve查询", "漏洞查询", "vulnerability lookup", "CVE编号", "查漏洞", "漏洞详情", "security advisory", "cve lookup", "漏洞缓解", "漏洞防御", "漏洞修复", "mitigation", "remediation", "概念性POC", "概念性PoC", "漏洞分析", "深度分析", "漏洞复现", "漏洞数据库", "构建漏洞库", "vulnerability database", "CVE数据库", "漏洞库查询", "数据库统计".

PPT-style web page builder with unified theming, slide management, element animations, and slide transitions. Creates self-contained HTML presentations with custom theme support. Enforces strict modification rules for project stability. Triggered by: "制作PPT", "PPT网页", "幻灯片", "演示页面", "presentation", "slide deck", "slide builder", "PPT风格", "添加动画", "页面过渡", "翻页动画", "新建幻灯片", "修改幻灯片", "删除幻灯片", "演示文稿", "slides", "ppt builder", "slide animation", "slide transition", "添加图片", "插入图片", "图片链接", "图床", "本地图片", "加图片", "放图片", "add image".

Design-driven frontend UI page generator that produces distinctive, production-grade HTML/CSS/JS pages avoiding generic AI aesthetics. Integrates design direction framework (purpose, tone, constraints, differentiation) before coding. Creates responsive single-file HTML viewable directly in browser without build tools or npm. Triggered by: "生成前端UI", "写个网页", "创建页面", "做个界面", "生成一个页面", "HTML页面", "浏览器查看", "web UI", "frontend page", "仪表板", "表单页面", "产品页", "landing page", "生成网页", "设计风格", "网页设计", "前端设计", "UI设计", "landing page", "dashboard", "现代风格", "独特设计", "design system".

Deeply analyze project codebase to generate hyper-detailed reports on project architecture, operation principles, workflows, and AI substitution blueprints. Each report documents every component down to function-level precision, serving as the foundation for creating replacement AI Skills. Triggered by: "分析项目", "项目架构", "代码库分析", "项目原理", "工作流分析", "AI替代方案", "项目复盘", "codebase analysis", "project architecture review", "架构分析", "项目review", "arch分析", "运行原理", "AI自动化分析", "Skill蓝图", "项目函数级分析", "逐步骤分析".

Organize course materials based on course syllabi and outlines. Tailors content for different learning phases: pre-class preparation, after-class review, sprint organization, and targeted challenge. Adapts material depth, structure, and focus according to time available and learning goals. Produces structured, phase-appropriate study resources with learning plans, practice materials, and progress tracking. Triggered by: "课程资料", "课程大纲", "预习资料", "复习资料", "course materials", "syllabus", "根据大纲整理", "学习曲线", "整理课程", "复习计划", "速成", "复习重点", "知识点梳理", "从零开始学", "课程笔记", "学习路线", "教学大纲", "知识体系".

A pure-AI replacement for the entire DeepWiki pipeline: analyze code repositories, generate structured Wiki documentation, answer codebase questions via simulated RAG, perform deep multi-turn research, and export documentation — all using only Claude's built-in capabilities with zero external dependencies. No servers, no FAISS, no LLM providers, no Docker. Triggered by: "deepwiki", "wiki project", "analyze repo", "codebase wiki", "document code", "repo analysis", "代码分析", "项目文档", "仓库 wiki", "code Q&A", "deep research", "研究代码库".

Transform outlines, syllabi, or knowledge point lists into structured self-study materials. Breaks down topics into digestible learning units with explanations, examples, exercises, and review questions. Searches for supplementary resources to enrich understanding. Helps users independently master any subject. Triggered by: "自学", "学习大纲", "知识点", "自学资料", "学习指南", "self-study", "study guide", "学习计划", "自学指导", "整理学习资料", "自学材料".

Plan structured learning paths for beginners who want to learn new technologies or subjects. Assesses current level, decomposes topics into progressive stages, and generates comprehensive study plans with resource recommendations and practice exercises. Triggered by: "我想学", "学习路线", "怎么学", "从零开始", "入门", "学习计划", "study plan", "learning path", "how to learn", "getting started", "从零开始学", "学习规划", "新手".

Analyze any project in the working directory, understand its architecture, operating principles, data flow, and workflows, and automatically convert it into a reusable Claude Code Skill. If the project is unsuitable for skill conversion, inform the user with specific reasons and terminate. Triggered by: "项目转skill", "项目转技能", "从项目创建skill", "project to skill", "convert project to skill", "从代码生成skill", "根据项目生成skill", "代码转技能", "仓库生成skill", "项目自动生成skill", "codebase to skill", "repo to skill", "项目转化为技能", "仓库转化为skill".

Systematically optimize user-provided prompts with formal verification at every stage: quantitative 12-dimension scoring (0-5 scale), anti-pattern detection, cross-dimension dependency mapping, formal optimization primitives with pre/post-conditions, counterfactual reasoning, conflict resolution protocol, and regression verification (re-analyze optimized output through the same framework). Each optimization is grounded in explicit logical inference with measurable quality metrics — never arbitrary modification. Supports 12+ domain-specific strategies with composable optimization operations. Triggered by: "优化提示词", "优化prompt", "提示词优化", "prompt优化", "rewrite prompt", "improve prompt", "改进提示词", "prompt改写", "逻辑推断优化", "精准提示词", "领域优化", "优化指令", "prompt engineering", "提示词工程".