Agent Skills: Code Implementation Methodology

Code Implementation Methodology

This skill provides systematic approach to implementing code that blends naturally with existing codebase as if written by the original developers.

Leverages: [codebase-analysis] skill for discovering project conventions and patterns.

Core Implementation Principle

Mimic, Don't Import: Every line of new code should feel like it was written by the same developer who wrote the surrounding code.

Implementation Workflow

Phase 1: Context Learning

Before writing any code:

1. Study neighbors: Read 3-5 files in same directory/module
2. Analyze similar features: Find comparable functionality
3. Learn patterns: Identify repeated structures and approaches
4. Understand dependencies: See how libraries are used
5. Examine tests: Learn validation approaches

Phase 2: Pattern Matching

Match existing code exactly:

1. Naming: Use same case, prefixes, suffixes as neighbors
2. Structure: Follow same file organization and class/function layout
3. Style: Match indentation, spacing, bracing, comments
4. Libraries: Use same dependencies and utilities as existing code
5. Error handling: Follow same error patterns and logging

Phase 3: Implementation

Write code that blends in:

1. Copy structural patterns from similar existing code
2. Reuse existing utilities and helper functions
3. Follow same abstraction levels as surrounding code
4. Match comment style and documentation format
5. Use same testing patterns and assertions

Phase 4: Integration Validation

Ensure seamless integration:

1. Run existing tests - all must pass
2. Execute build commands - must compile/run
3. Verify style consistency with surrounding code
4. Check that integration points work correctly
5. Ensure no new dependencies without justification

Style Matching Categories

Naming Conventions

• Match case style (camelCase, snake_case, PascalCase)
• Follow prefix/suffix patterns (get*, is*, has*, *Service, *Repository)
• Use domain terminology from existing code
• Maintain consistency within same module

Structural Patterns

• Copy class/function organization from similar files
• Follow same constructor/initialization patterns
• Match method ordering (public → private, create → read → update → delete)
• Replicate abstraction levels

Import Organization

• Group imports same way as existing files
• Use same libraries and frameworks
• Follow import ordering conventions
• Avoid introducing new dependencies

Error Handling

• Match exception types and error messages
• Follow logging patterns and levels
• Use same error propagation approach
• Replicate validation patterns

Comment and Documentation

• Match comment style (// vs /* */ vs #)
• Follow documentation format (JSDoc, Python docstrings, etc.)
• Use same level of detail as surrounding code
• Maintain consistent spacing and formatting

Common Implementation Patterns

Adding CRUD Operations

When adding CRUD, find existing CRUD examples:

1. Copy the structure exactly
2. Adjust entity/model names
3. Keep same validation patterns
4. Use same repository patterns
5. Follow same endpoint/API structure

Adding API Endpoints

When adding endpoints, find similar endpoints:

1. Copy routing pattern
2. Match request/response DTOs structure
3. Use same validation approach
4. Follow same error response format
5. Apply same authentication/authorization

Adding Business Logic

When adding logic, study similar use cases:

1. Follow same service/use case structure
2. Use existing domain models
3. Apply same transaction patterns
4. Reuse existing validation
5. Match logging and error handling

Integration Best Practices

Use Existing Utilities

✅ Always: Reuse project utilities and helpers
❌ Never: Reimplement existing functionality

Follow Architectural Boundaries

✅ Always: Respect layer dependencies (API → Service → Repository)
❌ Never: Skip layers or create circular dependencies

Match Technology Choices

✅ Always: Use same libraries as existing code
❌ Never: Introduce new frameworks without strong justification

Quality Checklist

Before considering implementation complete:

• [ ] Does code match naming conventions of surrounding files?
• [ ] Are imports organized like existing files?
• [ ] Does error handling follow project patterns?
• [ ] Are same libraries/utilities used?
• [ ] Do comments match existing style?
• [ ] Does code structure align with similar features?
• [ ] All existing tests still pass?
• [ ] New code follows discovered patterns?
• [ ] No architectural boundaries violated?
• [ ] No unnecessary dependencies added?

Detailed Examples

For comprehensive code examples demonstrating each pattern, see:

• examples.md - Detailed code samples for all matching strategies

Remember: The best implementation is invisible - it looks like it was always part of the codebase. Use [codebase-analysis] to understand the codebase's voice, then write in that voice.

Mandatory: Tone & Reference Policy

• 모든 기술적 주장/판단에 출처 명시 (공식 문서 URL, GitHub issue, RFC 등)
• 출처 없는 주장은 "출처 미확인 — 검증 필요"로 표기
• "~일 수 있습니다", "아마도", "대체로" 등 hedging 표현 금지. 확인 상태를 명시.
• 톤: 간결, 사실 중심. 불필요한 수식어 배제.
• 추론이 아닌 실제 확인된 사실만으로 구현/분석 진행.

Mandatory: Code Principles

• TDD: 테스트 먼저 → 구현 → 리팩토링
• SSoT/DRY: 중복 정의 금지, 반복 패턴은 추상화 검토
• SRP: 하나의 모듈/함수는 하나의 책임
• DDD: 도메인 용어 반영, 경계 컨텍스트 존중
• Functional: 불변성 우선, 부수효과 격리

Mandatory: Integration Verification

구현 완료 후 반드시:

1. 빌드 성공 확인
2. 전체 테스트 통과 확인
3. lint/format 통과
4. 실제 동작 확인
5. 회귀 없음 확인

완료 보고 형식:

[COMPLETE] {task} — Tests: pass/fail, Build: ok/fail, Integration: verified