When to open this: You have heard of Claude Skills and want to see a complete example that is "actually built, published to NPM, and has the full SKILL.md / references / scripts trio". This wiki breaks down 3 Skills built by newtype (Huang Yihe), all of which are standard implementations of the Anthropic Skill format,and most importantly, how a Skill orchestrates the collaboration between "Sequential Thinking + framework library + MCP".
Relationship with newtype · MCP Protocol and Ecosystem in Practice relationship: That document covers the MCP protocol ecosystem, while this wiki covers the Skill layer of abstraction — Skills call MCP, Skills call Sequential Thinking, and Skills provide framework presets.
1. What is a Skill (Anthropic's perspective)
Skill = .skill file = zip package, containing:
SKILL.md(required): core definition of the Skill + workflowreferences/(optional): detailed documentation (framework explanations, usage examples, style guides)scripts/(optional): Python / TypeScript helper scriptsREADME.md / CHANGELOG.md / INSTALLATION-GUIDE.md(optional): user documentation
Frontmatter of SKILL.md:
---
name: super-analyst
description: 简短描述 + 触发条件(LLM 自动识别何时调用此 Skill)
---
Why Skill instead of MCP?
- MCP is for connecting external tools(GitHub / Notion / databases)
- Skill is for encapsulating Agent behavior patterns(how to think, how to search, how to make decisions)
- Skills can call MCP, not the other way around
newtype's judgment (see newtype · MCP Protocol and Ecosystem in Practice):
"Skill-ification is the end state. Once any prompt + workflow is stable enough, turn it into a Skill."
2. Skill #1: Super Analyst v1 (2025-Q3)
2.1 Core positioning
An intelligent selector for 12 analysis frameworks. User asks a question → the Skill automatically selects the most suitable framework (or combination) → retrieves the detailed prompt for that framework via MCP → executes systematically.
2.2 SKILL.md description (frontmatter)
name: super-analyst
description: Advanced analytical framework selector for complex problem-solving
and decision-making. Use when users need structured analysis, strategic thinking,
problem diagnosis, or decision support. Automatically selects and applies the
most appropriate framework(s) from 12 specialized analytical methods including
First Principles Thinking, 5 Whys, SWOT, Porter's Five Forces, Cost-Benefit
Analysis, Design Thinking, Systems Thinking, Socratic Method, Pareto Analysis,
MECE Principle, Hypothesis-Driven Analysis, and Scenario Planning.
2.3 The 12 analysis frameworks (full reference table)
| Number | Framework | Applicable to | When to use |
|---|---|---|---|
| 1 | First Principles Thinking | Innovation, breakthrough problems, complex decisions | When fundamental redesign is needed / multiple variables are uncertain / traditional methods fail |
| 2 | 5 Whys | Fault diagnosis, process improvement | Manufacturing failures, software bugs, customer complaints, recurring errors |
| 3 | SWOT Analysis | Business strategy, project planning | Assessing competitiveness, new businesses, pre-launch risks |
| 4 | Porter's Five Forces | Industry analysis, competitive assessment | Market entry decisions, industry investment, mergers and acquisitions, pricing |
| 5 | Cost-Benefit Analysis | Investment evaluation, policy decisions | Infrastructure / technology upgrades, comparison of alternative options |
| 6 | Design Thinking | Product/service innovation, user problems | Developing apps / UX, education and health challenges |
| 7 | Systems Thinking | Complex systems, macro-level decisions | Seeing overall interactions |
| 8 | Socratic Method | Exploring essence, debate | Revealing assumptions through questioning |
| 9 | Pareto Analysis (80/20) | Prioritization, resource allocation | 80% of outcomes come from 20% of causes |
| 10 | MECE Principle | Structured analysis (McKinsey) | Collectively exhaustive, mutually exclusive |
| 11 | Hypothesis-Driven Analysis | Scientific method | Hypothesize first, then verify |
| 12 | Scenario Planning | Uncertainty, long-term strategy | Multiple future scenarios |
2.4 Framework prompts are retrieved via MCP (key design)
"Retrieves the detailed framework prompt using prompt-house-local MCP server"
- The Skill itself does not store the 12 full prompts (that would require tens of thousands of words)
- Prompts are stored in prompt-house-local MCP inside
- Skill workflow:Analyze requirements → select framework → retrieve prompt via MCP → apply → output
This is newtype's design philosophy: A Skill is the routing + selection + orchestration layer; specific content (prompts) resides in MCP (and can be independently maintained and reused).
2.5 references files
references/framework-mapping.md: Detailed reference table mapping frameworks to applicable scenariosreferences/usage-examples.md: Real-world cases for each frameworkscripts/framework_selector.py: Python helper logic
3. Skill #2: Super Analyst Pro 2.0 (2025-10-26)
3.1 Core upgrades (vs v1)
| Dimension | v1 | v2.0 (Pro) |
|---|---|---|
| Line count | 180 | 539(3x) |
| Workflow | Simple selection | 7-Stage Systematic Workflow |
| Thinking | None | Dual Sequential Thinking(2 decision points) |
| Information | None | Web Search integration(Chinese-English bilingual) |
| Complexity | One-size-fits-all | Level 1/2/3 adaptive |
| Output | Framework result | Consulting-Grade Insights |
Slogan:Deep Thinking + Comprehensive Research + Structured Analysis = Consulting-Grade Insights
3.2 7-Stage workflow
Stage 1: Problem Understanding & Assessment (10-30秒)
↓
Stage 2: Intelligence Planning (Sequential Thinking #1) * [可跳]
↓
Stage 3: Intelligence Gathering [可跳]
↓
Stage 4: Framework Selection (Sequential Thinking #2) ** [必跑]
↓
Stage 5: Prompt Retrieval(via MCP)
↓
Stage 6: Structured Analysis
↓
Stage 7: Integrated Output
Key design:
- Stage 4 is the always-execute core — framework selection must be performed at any complexity level
- Stages 2-3 run depending on information needs (skipped for purely conceptual questions)
- Sequential Thinking is used twice(planning + decision-making) — this is newtype's core pattern
3.3 Level 1/2/3 complexity determination
| Level | Time | Description | Example |
|---|---|---|---|
| Level 1 Simple | 1-2 minutes | Concept explanation, framework description | "What is SWOT analysis?" |
| Level 2 Medium | 3-6 minutes | Partial external information + 2-3 dimensions | "Analyze Tesla's competitive advantages" |
| Level 3 Complex | 10-20 minutes | Deep research + multi-dimensional strategy | "Should we enter Indian market?" |
3.4 Information need Trigger table
✅ External information required (run Stages 2-3):
- Specific companies/products/markets/industries
- Current data (prices, rankings, trends)
- Cases, best practices, competitive intelligence
- Recent events
❌ Not required (skip to Stage 4):
- Pure concepts/theory
- User has already provided complete context
- General knowledge
- Framework explanations
3.5 Specific usage of Sequential Thinking at the two Decision Points
Decision Point #1(Stage 2)— Intelligence Planning:
- Use Sequential Thinking to break down: "What information do I need? Where should I look it up?"
- Output: Search Plan (4-6 search queries)
- Chinese-English bilingual parallel search (search Chinese for domestic information, English for international information)
Decision Point #2(Stage 4)— Framework Selection:
- Use Sequential Thinking to evaluate: Which framework(s) to use? Order? Weighting?
- Output: Framework Selection (execution plan for 1-3 frameworks)
3.6 Included file list
super-analyst-pro/
├── SKILL.md 539 行,主流程
├── README.md 用户文档,Quick Start
├── INSTALLATION-GUIDE.md 安装步骤
├── CHANGELOG.md v1 → v2.0 升级日志
├── UPGRADE-SUMMARY.md 升级亮点
└── references/
├── framework-mapping.md
├── search-strategy-guide.md
└── sequential-thinking-prompts.md
Comparison v1:v2.0 is a true "product" with complete product documentation. v1 is a "prototype".
4. Skill #3: Super Writer (2025-10-27)
4.1 Core positioning
Professional content creation assistant, with 3 major capabilities:
- 📚 Material Collection — intelligent web search + user material organization
- 🎨 Style Analysis — 7 style analysis methods to extract writing style from reference content
- ✍️ Content Writing — 6 proven writing methodologies
4.2 Trigger conditions (automatically recognized by LLM)
"Triggers when users need to create articles, marketing copy, stories, reports, social media posts, emails, blogs, or any text-based content creation tasks."
4.3 Key design:Triple Sequential Thinking
Stage 2: Material Planning(Sequential Thinking #1)
Stage 4.5: Style Analysis Planning(Sequential Thinking #2)
Stage 6: Methodology Selection(Sequential Thinking #3)
super-analyst is dual (double Sequential Thinking), super-writer is triple — more complex tasks require more thinking decision points.
4.4 7-Stage workflow
Stage 1: Content Request Analysis
↓
Stage 2: Material Planning(Sequential Thinking #1)*
↓
Stage 3: Material Collection
↓
Stage 4: Style Reference Check **
↓ [if reference]
Stage 4.5: Style Analysis Planning(Sequential Thinking #2)
↓
Stage 5: Style Analysis Execution
↓
Stage 6: Methodology Selection(Sequential Thinking #3)
↓
Stage 7: Content Writing + Output
4.5 7 Style Analysis methods
(from references/style-analysis-guide.md)
- Structural Analysis — paragraph organization, opening/closing patterns
- Tone Analysis — formal / casual / authoritative / conversational
- Vocabulary Profile — high-frequency words, sentence length distribution, jargon density
- Rhythm & Pacing — alternation of short and long sentences, rhythm patterns
- Voice Analysis — first / second / third person preference
- Persuasion Patterns — argumentation methods, emotional arc
- Cultural Markers — cultural references, allusions, local context
4.6 6 Writing Methodologies
(from references/methodology-mapping.md)
| Methodology | Applicable to |
|---|---|
| AIDA(Attention-Interest-Desire-Action) | Marketing copy |
| PAS(Problem-Agitation-Solution) | Sales pages |
| Pyramid Principle | Reports/analysis |
| Inverted Pyramid | News |
| Hero's Journey | Stories |
| Story Spine | Short narratives |
4.7 Key Format Rules (SKILL.md warnings)
🚨 CRITICAL: NEVER display in user-facing output:
❌ HTML tags like <details>, <summary>
❌ Raw tool call syntax
❌ Sequential Thinking tool output (process is internal)
✅ ALWAYS use clean Markdown
✅ Users see ONLY the RESULT of thinking, not the process
This reflects newtype's engineering judgment:Sequential Thinking is the Agent's internal state and should not be exposed to the user.
5. Common patterns across the 3 Skills (advanced Patterns)
5.1 5 major commonalities in Skill design
- Sequential Thinking is the core — instead of generating directly, think first then act
- Retrieve large blocks of content (prompts / templates) via MCP — the Skill itself stays lightweight, with content externalized
- 7-Stage workflow — all use 7 steps:understand → plan → collect → decide → execute → integrate → output
- Complexity adaptation(Level 1/2/3) — don't over-engineer simple tasks
- Complete productization(README + CHANGELOG + Installation guide)
5.2 The "router + executor" philosophy of Skills
User Request
↓
Skill 作为 "router"(路由器)
↓
分析 → 选最优 framework / methodology / strategy
↓
MCP / Sub-tool 作为 "executor"(执行器)
↓
取详细 prompt → 实际生成
A Skill does not solve specific problems; a Skill solves the meta-problem of "which method should be used".
5.3 Trigger condition design
All 3 Skills are triggered via the SKILL.md frontmatter's description:
- Be as specific as possible(list all 12 framework names, 7 style methods)
- Include use case keywords(automatically matched by the LLM)
- Include trigger sentences("Triggers when users need to...")
6. Takeaways for independent developers
6.1 When you should build a Skill (instead of an MCP / Workflow)
| Your need | Recommendation |
|---|---|
| Calling external APIs / tools (GitHub / Slack / DB) | MCP |
| You want to solidify an "analysis methodology / writing methodology / SOP" | Skill |
| You want to build a complete multi-step business process | Workflow(n8n / Dify) |
| You want to orchestrate "think first → then search → then decide → then execute" | Skill + Sequential Thinking |
6.2 6 suggestions for Skill engineering (from the 3 examples)
- Write frontmatter carefully: the LLM relies on this to decide whether to trigger
- 7-Stage is the sweet spot: fewer steps lack rigor, more steps are hard for users/LLMs to follow
- Use Sequential Thinking 2-3 times: at key decision points, not every step
- Don't stuff large prompts into SKILL.md: put them in references/ or MCP
- Write a CHANGELOG: make v1 → v2 upgrades clear
- Hide internal state: do not show Sequential Thinking output to users
6.3 How to deploy newtype's Skills to your own Claude
# 把 .skill 文件放到 Claude 的 skills 目录
~/.claude/skills/super-analyst.skill
# Claude 重启后,自动加载
# 通过 Slash command 显式调用或自然语言触发
newtype's judgment in newtype · MCP Protocol and Ecosystem in Practice has been validated:
"Skill-ification is the end state. Once built, it can be used repeatedly, across sessions, and by teams."
7. Links to other wikis
- newtype · MCP Protocol and Ecosystem in Practice — MCP protocol ecosystem (the tool layer for Skills)
- newtype · Prompt / Context Engineering in Practice — Prompt engineering (the prompt layer for Skills)
- Claude Code Best Practices (Cal Rueb · Anthropic Internal Talk) — Claude Code from Cal's perspective at Anthropic (this wiki presents newtype's self-built Skill examples)
- newtype · Vibe Coding Independent Product Practice — newtype's Independent Product Development History (Super Analyst / Super Writer are among these products)
- Full Evolution of newtype Knowledge System and Life OS — newtype OS Knowledge System (Skills are the "application" layer of the OS)
- Wang Kaiduo Agent Browser Matrix Architecture (v1.1 Practical Guide) — Wang Kaiduo Browser Agent Matrix (His Skill system follows a different paradigm from newtype)
History
- 2026-05-17:Phase 5.1 ingest Extract from three zip packages: file_023(Super Analyst v1.skill), file_024(Super Analyst Pro 2.0.skill), and file_025(Super Writer zip), merge and write 1 wiki. Fully cover SKILL.md / references / scripts / README / CHANGELOG of the 3 Skills.