A curated collection of reusable, structured AI skills — each is a complete methodology + prompt + workflow that helps LLMs perform complex tasks more effectively.
Not just prompt snippets. Each skill is a battle-tested playbook with clear structure, ready to plug into any LLM workflow.
Designed for:
- Claude / ChatGPT / Gemini / Open-source LLMs
- AI Agents & Copilot-style assistants
- Developers and knowledge workers
| Prompt Collections | AI Skill Kit | |
|---|---|---|
| Structure | One-liner prompts | Methodology + framework + examples |
| Reusability | Copy-paste | Modular, plug into any workflow |
| Quality control | Community contributed | Each skill tested and iterated |
| Scope | Generic tasks | Domain-specific deep skills |
ai-skill-kit/
├── ai-engineering/ # RAG, Agent, Prompt, LLM debugging
├── career/ # Interview, resume, capability profiling
├── analysis/ # Decision making, product strategy, deep analysis
├── learning/ # Structured learning paths, topic analysis
├── creative/ # Writing, illustration, SVG generation
├── development/ # Testing, evaluation, debugging
└── workflow/ # Automation, statistics, weekly review
| Skill | Description |
|---|---|
| rag-evaluator | Diagnose RAG pipeline issues across retrieval, generation, and consistency layers |
| prompt-optimizer | Identify root causes of prompt failures and apply structured fixes |
| agent-designer | Design ReAct / Plan-Execute / Multi-Agent architectures with failure handling |
| llm-debugger | Debug LLM apps in production: API errors, rate limits, token overflow, streaming |
| vector-db-guide | Vector DB selection and usage: Chroma, Milvus, pgvector, Qdrant compared |
| langchain-patterns | Core LangChain patterns: LCEL, RAG chain, memory, agents with real code |
| ai-solution-designer | Design AI solutions with scenario evaluation, architecture, risk and ROI analysis |
| Skill | Description |
|---|---|
| mock-interview | Generate high-value interview questions with answer frameworks based on resume + JD |
| job-seeker-resume-cn | Chinese resume optimization for general job market with platform algorithm strategies |
| ai-job-transition-resume | Resume optimization for AI engineer / AI pre-sales roles |
| ai-engineer-interviewer | Conduct technical interviews for AI engineering positions with role-specific evaluation |
| interview-sparring | Diagnostic interview practice with dynamic weak-point identification and coaching |
| capability-miner | Extract structured capability profiles from project memory and Git history |
| Skill | Description |
|---|---|
| product-analysis | Multi-dimensional product strategy analysis (user / solution / business / competition / risk) |
| decision-tree | Structured decision making with explicit recommendations |
| deep-analysis | Multi-perspective Self-Debate: generate, critique, refine, loops until score >= 8 |
| Skill | Description |
|---|---|
| learn-anything | Structured learning path for any technology or skill |
| atdf-analyzer | Systematically analyze AI topics using 8-dimensional ATDF framework |
| Skill | Description |
|---|---|
| article-illustrator | SVG illustrations for articles and visual communication |
| article-writer | Structured article generation from notes, outlines, or rough content |
| svg-generator | User-goal-first SVG generation workflow |
| svg-to-png | Reliable SVG to PNG conversion with tool fallbacks |
| Skill | Description |
|---|---|
| add-tests | Add unit tests to existing code, identify testable functions, and ensure all tests pass |
| eval-debug | Analyze interview JSONL files for model quality and program logic debugging |
| rag-eval | Analyze RAG experimental results, interpret metrics, and identify anomalies |
| Skill | Description |
|---|---|
| cc-stats | Generate Claude Code usage statistics and productivity insights |
| claude-scheduler | Schedule Claude Code tasks using macOS launchd for persistent background execution |
| retro | Analyze recent conversations to identify repeated operations for automation |
| weekly-output | Synthesize weekly notes using BASB, Zettelkasten, T-shape, and GTD frameworks |
Each skill folder contains a SKILL.md with the complete prompt and methodology. To use a skill:
- Open the
SKILL.mdfile in the skill folder - Copy the prompt into your LLM conversation
- Follow the structured workflow
MIT