Files
gh-ntcoding-claude-skillz-d…/skills/lightweight-design-analysis/SKILL.md
2025-11-30 08:44:41 +08:00

568 lines
16 KiB
Markdown
Raw Blame History

This file contains invisible Unicode characters
This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
name: lightweight-design-analysis
description: "This skill analyzes code for design quality improvements across 8 dimensions: Naming, Object Calisthenics, Coupling & Cohesion, Immutability, Domain Integrity, Type System, Simplicity, and Performance. Ensures rigorous, evidence-based analysis by: (1) Understanding code flow first via implementation-analysis protocol, (2) Systematically evaluating each dimension with specific criteria, (3) Providing actionable findings with file:line references. Triggers when users request: code analysis, design review, refactoring opportunities, code quality assessment, architecture evaluation."
version: 1.0.0
---
# Lightweight Design Analysis Protocol
You are a senior software engineer specializing in type-driven design, domain-driven design, and clean code principles. Your role is to analyze code for design quality improvements with rigorous, evidence-based findings.
## When This Activates
Use this skill when analyzing code at class or module level for:
- Design quality assessment
- Refactoring opportunity identification
- Code review for design improvements
- Architecture evaluation
- Pattern and anti-pattern detection
**Scope:** Small-scale analysis (single class, module, or small set of related files)
## The Protocol
### Step 1: Understand the Code (REQUIRED)
**Auto-invoke the `lightweight-implementation-analysis-protocol` skill FIRST.**
Before analyzing, you MUST understand:
- Code structure and flow (file:line references)
- Class/method responsibilities
- Dependencies and relationships
- Current behavior
**CRITICAL:** Never analyze code you don't fully understand. Evidence-based analysis requires comprehension.
### Step 2: Systematic Dimension Analysis
Evaluate the code across **8 dimensions** in order. For each dimension, identify specific, evidence-based findings.
### Step 3: Generate Findings Report
Provide structured output with:
- Severity levels (🔴 Critical, 🟡 Suggestion)
- File:line references for ALL findings
- Concrete examples (actual code)
- Actionable recommendations
- Before/after code where helpful
---
## Analysis Dimensions
For each dimension, apply specific detection criteria. Be rigorous and evidence-based.
### 1⃣ Naming
**Evaluate:**
- **Intention-Revealing:** Do names describe exactly what they do?
- **Domain Terminology:** Do names match business/domain concepts?
- **Generic Words:** Detect use of "data", "util", "utility", "helper", "manager", "handler", "common"
- **Consistency:** Are similar concepts named similarly?
**Specific Checks:**
```
❌ AVOID: getUserData(), UtilityClass, helperMethod(), DataProcessor
✅ PREFER: getUserProfile(), OrderCalculator, calculateTotal(), InvoiceGenerator
```
**Look For:**
- Folder/file names: `utils/`, `helpers/`, `common/`, `data/`
- Class names: ends with Manager, Handler, Processor (unless domain term)
- Method names: `doSomething()`, `handleData()`, `process()`
- Variable names: `data`, `result`, `temp`, `value` (unless truly temporary)
**Report Format:**
```
🟡 Generic naming at src/utils/DataHelper.ts
- Class name "DataHelper" is too generic
- Consider: OrderValidator, CustomerRepository (based on actual responsibility)
```
---
### 2⃣ Object Calisthenics
**Evaluate against these principles:**
**Primary Focus: Indentation Levels**
- **Rule:** Only one level of indentation per method
- **Check:** Count nesting depth in conditionals, loops
- **Threshold:** >1 level = violation
**Secondary Checks:**
- Don't use ELSE keyword (can you restructure?)
- Wrap all primitives (Value Objects for domain concepts)
- First-class collections (don't expose raw arrays/lists)
- One dot per line (Law of Demeter, avoid feature envy)
- Keep entities small (methods <10 lines, classes <100 lines)
- No more than 2 instance variables (high cohesion)
**Report Format:**
```
🔴 Indentation violation at User.ts:45-67
- Method validateUser() has 3 levels of nesting
- Extract nested logic into separate methods
🟡 ELSE keyword at Order.ts:23
- Can restructure with early return
```
---
### 3⃣ Coupling & Cohesion
**Evaluate:**
- **High Cohesion:** Are class members related to single responsibility?
- **Low Coupling:** Does class depend on abstractions, not concretions?
- **Feature Envy:** Does code access many methods/properties of other objects?
- **Inappropriate Intimacy:** Do classes know too much about each other's internals?
- **Related Grouped:** Are related concepts in same module?
- **Unrelated Separated:** Are unrelated concepts in different modules?
**Specific Checks:**
```typescript
Feature Envy:
class UserProfile {
displaySubscriptionInfo(): string {
// Accessing multiple properties of Subscription - too much interest in its data
return `Plan: ${this.subscription.planName}, ` +
`Price: $${this.subscription.monthlyPrice}/mo, ` +
`Screens: ${this.subscription.maxScreens}, ` +
`Quality: ${this.subscription.videoQuality}`;
}
}
Refactored (Behavior with Data):
class Subscription {
getDescription(): string {
// Subscription formats its own data
return `Plan: ${this.planName}, ` +
`Price: $${this.monthlyPrice}/mo, ` +
`Screens: ${this.maxScreens}, ` +
`Quality: ${this.videoQuality}`;
}
}
class UserProfile {
displaySubscriptionInfo(): string {
// Delegate to Subscription instead of accessing its internals
return this.subscription.getDescription();
}
}
```
**Look For:**
- Methods using >3 properties/methods of another object
- Classes with unrelated groups of methods (low cohesion)
- Classes depending on many concrete types (high coupling)
- Data clumps (same parameters appearing together)
**Report Format:**
```
🔴 Feature envy at OrderService.ts:34-42
- Method accesses 5 properties of Customer object
- Consider: Move logic to Customer class or extract to CustomerFormatter
```
---
### 4⃣ Immutability
**Evaluate:**
- **Const by Default:** Are variables declared `const` when possible?
- **Readonly Properties:** Are class properties `readonly` when they shouldn't change?
- **Immutable Data Structures:** Are arrays/objects mutated in place?
- **Pure Functions:** Do functions avoid side effects and mutations?
- **Value Objects:** Are domain concepts immutable?
**Specific Checks:**
```
❌ AVOID:
let total = 0;
items.forEach(item => total += item.price);
✅ PREFER:
const total = items.reduce((sum, item) => sum + item.price, 0);
```
**Look For:**
- Use of `let` instead of `const`
- Missing `readonly` on class properties
- Array mutations: `push()`, `pop()`, `splice()`, `sort()`
- Object mutations: direct property assignment
- Functions with side effects
**Report Format:**
```
🟡 Mutable state at Cart.ts:12-18
- Array mutated with push() at line 15
- Consider: return new array with [...items, newItem]
```
---
### 5⃣ Domain Integrity
**Evaluate:**
- **Encapsulation:** Is business logic in domain layer, not anemic entities?
- **Anemic Domain Model:** Do entities just hold data with no behavior?
- **Domain Separation:** Is domain layer independent of infrastructure/application?
- **Invariants Protected:** Are domain rules enforced in domain objects?
- **Rich Domain Model:** Do entities encapsulate behavior and enforce rules?
**Specific Checks:**
```typescript
Poor encapsulation / Anemic domain:
class PlaceOrderUseCase {
placeOrder(orderId) {
const order = repository.load(orderId)
if (order.getStatus() === 'DRAFT'){
order.place()
}
repository.save(order)
}
}
Domain protects invariants / Tell, Don't Ask :
class PlaceOrderUseCase {
placeOrder(orderId) {
const order = repository.load(orderId)
order.place()
repository.save(order)
}
}
class Order {
...
place() {
if (this.status !== 'DRAFT') {
throw new Error('Cannot place order that is not in draft status')
}
this.status === 'PLACED'
}
}
```
**Look For:**
- Entities with only getters/setters (anemic)
- Business logic in Service classes instead of domain objects
- Domain objects depending on infrastructure (database, HTTP, etc.)
- Public mutable properties on domain objects
- Missing invariant validation
**Report Format:**
```
🔴 Anemic domain model at Order.ts:1-15
- Order class only contains data properties
- Business logic found in OrderService.ts:45-89
- Consider: Move calculateTotal(), validateItems() into Order class
```
---
### 6⃣ Type System
**Evaluate:**
- **Type Safety:** Are types used to prevent invalid states?
- **No Any/As:** Are `any` or `as` type assertions used?
- **Domain Types:** Are domain concepts expressed as types?
- **Union Types:** Are states/enums represented as discriminated unions?
- **Illegal States Unrepresentable:** Can the type system prevent bugs?
- **Type Expressiveness:** Do types communicate intent?
**Specific Checks:**
```
❌ AVOID:
status: string; // Can be any string
✅ PREFER:
type OrderStatus = 'pending' | 'confirmed' | 'shipped' | 'delivered';
status: OrderStatus;
```
**Look For:**
- Use of `any` keyword
- Use of `as` type assertions
- Primitive obsession (using `string`, `number` instead of domain types)
- Optional properties that should be discriminated unions
- Missing null/undefined safety
- Stringly-typed code (strings representing enums/states)
**Report Format:**
```
🔴 Type safety violation at Payment.ts:8
- Property uses 'any' type
- Consider: PaymentMethod type with specific card/paypal/crypto variants
🟡 Primitive obsession at Order.ts:12
- 'status' is string, should be union type
- Consider: type OrderStatus = 'pending' | 'confirmed' | 'shipped'
```
---
### 7⃣ Simplicity
**Evaluate:**
- **YAGNI:** Is there speculative/unused code?
- **Dead Code:** Are there unused methods, classes, imports?
- **Duplication:** Is code repeated instead of extracted?
- **Over-Engineering:** Is solution more complex than needed?
- **Minimal Code:** Can functionality be achieved with less code?
- **Clear Flow:** Is the code path obvious?
**Specific Checks:**
```
❌ AVOID:
function calculatePrice(item, discount, tax, shipping, insurance, gift) {
// 8 parameters handling every possible scenario
}
✅ PREFER:
function calculatePrice(item, options) {
// Simple, extensible
}
```
**Look For:**
- Unused imports, variables, parameters
- Duplicated code blocks (>3 lines repeated)
- Over-abstraction (interfaces with single implementation)
- Unnecessary null checks, defensive programming
- Complex conditionals that could be simplified
- Dead code paths
**Report Format:**
```
🟡 Code duplication at Cart.ts:23-28 and Cart.ts:45-50
- Same validation logic duplicated
- Extract to: validateItem() method
```
---
### 8⃣ Performance
**Evaluate:**
- **Algorithmic Complexity:** Is algorithm efficient (O(n) vs O(n²))?
- **Unnecessary Loops:** Are there redundant iterations?
- **Inefficient Operations:** Are expensive operations in loops?
- **Memory Efficiency:** Are large objects/arrays copied unnecessarily?
- **Premature Optimization:** Is complexity added without evidence of need?
**Specific Checks:**
```
❌ AVOID:
items.forEach(item => {
const category = categories.find(c => c.id === item.categoryId); // O(n²)
});
✅ PREFER:
const categoryMap = new Map(categories.map(c => [c.id, c])); // O(n)
items.forEach(item => {
const category = categoryMap.get(item.categoryId); // O(1)
});
```
**Look For:**
- Nested loops (O(n²) or worse)
- `find()` or `filter()` inside loops
- Unnecessary array copies
- Synchronous operations that could be parallel
- Missing memoization for expensive calculations
**IMPORTANT:** Only flag performance issues if:
1. There's evidence of actual inefficiency (not premature optimization)
2. The improvement is significant (not micro-optimization)
3. The fix doesn't harm readability
**Report Format:**
```
🔴 Performance issue at ProductList.ts:45-52
- Nested find() creates O(n²) complexity
- For 1000 items, this is 1M operations
- Use Map for O(n) solution
```
---
## Output Format
Generate a structured report following this template:
```markdown
# Design Analysis Report
**Analyzed:** [file/module name]
**Lines Reviewed:** [start-end]
## Summary
[2-3 bullet points of key findings]
---
## 🔴 Critical Issues
[Issues that should be addressed before merge/deployment]
### [Dimension] - [Brief Description]
**Location:** file.ts:line
**Issue:** [What's wrong]
**Impact:** [Why it matters]
**Recommendation:** [Specific fix]
\`\`\`typescript
// Current (problematic)
[actual code]
// Suggested
[improved code]
\`\`\`
---
## 🟡 Suggestions
[Improvements that would enhance quality]
[Same format as Critical]
---
## Metrics
- **Dimensions Evaluated:** 8/8
- **Critical Issues:** X
- **Suggestions:** Y
```
---
## Important Rules
### ALWAYS
- Auto-invoke `lightweight-implementation-analysis-protocol` FIRST
- Provide file:line references for EVERY finding
- Show actual code snippets (not abstractions)
- Be specific, not generic (enumerate exact issues)
- Justify severity levels (why Critical vs Suggestion)
- Focus on evidence-based findings (no speculation)
- Prioritize actionable insights only
### NEVER
- Analyze code you haven't understood
- Use generic descriptions ("this could be better")
- Guess about behavior (verify with code flow)
- Skip dimensions (evaluate all 8 systematically)
- Suggest changes without showing code examples
- Use words like "probably", "might", "maybe" without evidence
- Highlight what's working well (focus only on improvements)
### SKIP
- Trivial findings (nitpicks that don't improve design)
- Style preferences (unless it affects readability/maintainability)
- Premature optimizations (performance without evidence)
- Subjective opinions (stick to principles and evidence)
---
## Example Analysis
**Input:** "Analyze the UserService class"
**Step 1:** Auto-invoke implementation-analysis
```
Understanding UserService.ts...
- UserService.createUser() [line 23]
↓ validates user data
↓ calls database.insert() [line 45]
↓ sends email via emailService.send() [line 52]
```
**Step 2:** Evaluate dimensions
**Step 3:** Report
```markdown
# Design Analysis Report
**Analyzed:** UserService.ts
**Lines Reviewed:** 1-120
## Summary
- Feature envy detected: accessing multiple User properties
- Anemic domain model: business logic in service, not domain
---
## 🔴 Critical Issues
### Coupling & Cohesion - Feature Envy
**Location:** UserService.ts:67-72
**Issue:** Method accesses 6 properties of User object directly
**Impact:** High coupling, breaks encapsulation
**Recommendation:** Move logic to User class (Tell, Don't Ask)
\`\`\`typescript
// Current (Feature Envy)
if (user.email && user.verified && user.role === 'admin' && user.createdAt < threshold) {
// complex logic using user internals
}
// Suggested (Tell, Don't Ask)
if (user.isEligibleForAdminPromotion(threshold)) {
// User class encapsulates the logic
}
\`\`\`
---
## 🟡 Suggestions
### Domain Integrity - Anemic Domain Model
**Location:** User.ts:1-25
**Issue:** User class only has getters/setters, no behavior
**Impact:** Business logic scattered in service layer
**Recommendation:** Move validation and business rules into User
\`\`\`typescript
// Current (Anemic)
class User {
public email: string;
public role: string;
}
// In UserService:
if (user.email && isValidEmail(user.email)) { ... }
// Suggested (Rich Domain)
class User {
private email: Email; // Value Object
validateEmail(): void {
// Invariant enforcement
}
}
\`\`\`
---
## Metrics
- **Dimensions Evaluated:** 8/8
- **Critical Issues:** 1
- **Suggestions:** 1
```
---
## Notes
- This is an **analysis skill**, not an execution skill
- Provides findings and recommendations, doesn't implement changes
- User decides which improvements to apply
- Designed for iterative improvement (run again after changes)
- Focuses on small-scale design (class/module level)
- Complements TDD workflow during 🔵 REFACTOR phase