Methodology Selector

An AI-powered skill that evaluates 34 software engineering methodologies against your project context, resolves philosophical contradictions, and generates a tailored engineering methodology guide.

What It Does

You describe a task, project, or problem. Methodology Selector:

1. Detects context automatically — project phase, team style, risk tolerance, innovation level — using a signal-based keyword scanning system
2. Scores 34 methodologies across 6 weighted dimensions
3. Resolves contradictions — when conflicting methodologies both score high (e.g., First Principles vs Pattern Thinking), it uses the detected context to pick the right one
4. Generates a complete methodology guide — not a list, but an integrated framework with concrete actions, artifact checklists, anti-patterns, and next steps

Quick Start

Prerequisites

• Claude Code or OpenCode with Claude Skills support
• A project or task description

Installation

Copy the .claude/skills/methodology-selector/ directory into your project's .claude/skills/ directory:

cp -r MethodologySelector/.claude/skills/methodology-selector /your-project/.claude/skills/

Or symlink it globally if your CLI supports it.

Usage

Just describe your task in natural language. The skill triggers on phrases like:

• "选方法论" / "用什么方法" / "methodology" / "工程方法"
• "怎么做这个项目" / "给我一个方法论" / "技术方案的方法论"
• Any task description where you want systematic engineering guidance

Example:

我们要给一个金融支付系统做架构评审，现有系统经常出线上故障，团队5个人，时间比较紧

The skill will:

1. Detect context: Risk=Low (金融), Phase=Architecture (架构评审), Innovation=Incremental (现有系统)
2. Score all 34 methodologies with 6-dimensional weighted scoring
3. Resolve contradictions based on detected context
4. Generate a methodology-guide.md with the full framework

How It Works

User Input
   │
   ▼
┌─────────────────────────────────────┐
│ Step 0: Context Detection           │  Signal-based keyword scanning
│  (signal-tables.md)                 │  80+ bilingual keywords, deterministic
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ Step 1: Understand Requirements     │  What/Why/Where/Who/Constraints
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ Step 2: 6-Dimension Scoring         │  34 methodologies × 6 dimensions
│  Problem Fit (28%)                  │  Weighted average → final score
│  Phase Fit (24%)                    │
│  Scale Fit (18%)                    │
│  Domain Fit (14%)                   │
│  Composability (8%)                 │
│  Context Consistency (8%)           │
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ Step 3: Select (>= 85%)             │  Filter, rank, identify interactions
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ Step 3.5: Resolve Tensions          │  Context-driven contradiction resolution
│  6 contradiction pairs              │  Transparent reasoning in output
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ Step 4: Generate Guide              │  8-section structured Markdown document
│  (output-template.md)               │  Mermaid diagrams, artifact checklists
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ Step 5: Save Document               │  methodology-guide.md
└─────────────────────────────────────┘

Context Detection

Instead of relying on LLM intuition, context is detected through deterministic signal scanning:

Dimension	Values	Example Signals
Risk Tolerance	Low / Medium / High	金融/医疗 → Low; MVP/创业 → High
Innovation Level	Disruptive / Incremental / Pattern Reuse	从零开始 → Disruptive; 重构/遗留 → Incremental
Team Style	Experience-driven / Theory-driven / Pragmatic / Data-driven	TDD/CI → Data-driven; 能跑就行 → Pragmatic
Project Phase	Requirements / Architecture / Implementation / Testing / Diagnosis / Cost Estimation	架构评审 → Architecture; 排查故障 → Diagnosis

See references/signal-tables.md for the full keyword library (80+ signals, bilingual).

Contradiction Resolution

When contradictory methodologies both score >= 85%, the system resolves them based on context:

Contradiction	Resolution Rule
First Principles vs Pattern Thinking	Disruptive → First Principles; Pattern Reuse → Pattern Thinking
Reductionism vs Systems Thinking	Single-module → Reductionism; Cross-module → Systems Thinking
YAGNI vs Second-Order Thinking	High risk tolerance → YAGNI; Low risk tolerance → Second-Order
Pragmatism vs Formal Methods	High risk tolerance → Pragmatism; Low risk tolerance → Formal Methods
Dialectics vs Formal Logic	Pragmatic team → Dialectics; Theory-driven team → Formal Logic
Bayesian Thinking vs First Principles	Disruptive → First Principles; Otherwise → Bayesian Thinking

Methodology Library

34 methodologies across 7 layers:

Layer	Count	Examples
Thinking Paradigms	4	First Principles, Systems Thinking, Reductionism, Critical Thinking
Engineering Methods	5	Divide & Conquer, Abstraction, Design by Contract, SoC, Feedback-Driven
Process Frameworks	4	Agile, Lean, DevOps, DDD
Quality Assurance	4	TDD, Formal Methods, Chaos Engineering, Threat Modeling
Decision Methods	4	YAGNI, KISS, Trade-off Analysis, Occam's Razor
Philosophical Foundations	7	Ontology, Epistemology, Pragmatism, Formal Logic, Phenomenology, Ethics, Dialectics
Advanced Thinking Frameworks	6	Inversion, Second-Order Thinking, Theory of Constraints, Bayesian Thinking, Game Theory, Pattern Thinking

See references/methodology-library.md for the complete catalog.

Output Structure

The generated guide contains 8 sections:

Section	Content
0. Context Detection	Detected context + methodology tension resolutions
1. Requirements Summary	Core requirements linked to context
2. Selection Results	Full scoring breakdown, interaction map (Mermaid), rejected methodologies
3. Integrated Framework	Phase-by-phase guidance synthesized from selected methodologies
4. Artifacts Checklist	All deliverables with acceptance criteria
5. Anti-Patterns	Context-specific pitfalls to avoid
6. Applicability Boundaries	When to stop using each methodology
7. Verification Framework	Leading/lagging indicators, review checkpoints
8. Next Actions	Immediate prioritized action items

Project Structure

.claude/skills/methodology-selector/
├── SKILL.md                          # Main skill definition and process
└── references/
    ├── methodology-library.md        # 34 methodologies catalog (M01-M34)
    ├── signal-tables.md              # Context detection keyword library
    └── output-template.md            # Output document structure
README.md
LICENSE
.gitignore

Extending

Add a New Methodology

1. Add an entry to references/methodology-library.md with the next M number
2. If it contradicts an existing methodology, add a resolution rule to SKILL.md Step 3.5
3. Update the methodology count in SKILL.md line 14

Add Context Signals

Edit references/signal-tables.md. Each signal has: keyword(s), direction, confidence level. The detection rules in SKILL.md Step 0 will automatically pick up new signals.

Modify Scoring Weights

Edit the scoring dimensions table in SKILL.md Step 2. Weights must sum to 100%.

Customize Output

Edit references/output-template.md. The generated document follows this template exactly.

Contributing

See CONTRIBUTING.md for guidelines on adding methodologies, signals, or improving the scoring system.

License

MIT — see LICENSE.