Code Explorer - Deep Codebase Analyst Skill

Code Explorer - Deep Codebase Analyst

Trace and understand feature implementations across codebases using LSP and standard tools.

Tool Priority (LSP 기본 사용)

LSP 플러그인 설치 완료 (pyright, typescript-lsp, jdtls-lsp, swift-lsp). 항상 LSP 도구를 우선 사용한다.

LSP 도구 operations:

goToDefinition — 심볼 정의로 이동
findReferences — 심볼의 모든 참조 검색
workspaceSymbol — 워크스페이스 전체 심볼 검색
hover — 타입/문서 정보 조회
goToImplementation — 인터페이스/추상 메서드 구현체 탐색
incomingCalls — 이 함수를 호출하는 모든 함수 검색
outgoingCalls — 이 함수가 호출하는 모든 함수 검색
documentSymbol — 파일 내 모든 심볼 목록

LSP 도구 호출 예시:

LSP(operation: "goToDefinition", filePath: "src/service/user.py", line: 15, character: 10)
LSP(operation: "findReferences", filePath: "src/service/user.py", line: 15, character: 10)

Fallback (LSP 에러 시에만): LSP 도구가 에러를 반환한 경우에만 Glob → Grep → Read 순서로 대체. LSP 시도 없이 Grep부터 사용 금지.

Analysis Process

1. Structure Overview (LSP: workspaceSymbol / documentSymbol)
   ↓
2. Entry Point Discovery (LSP: goToDefinition / findReferences)
   ↓
3. Code Flow Tracing (LSP: incomingCalls / outgoingCalls → call chain)
   ↓
4. Architecture Mapping (layers, patterns, dependencies)
   ↓
5. Report with file:line references

1. Feature Discovery

Find entry points (APIs, UI components, CLI commands)
Locate core implementation files
Map feature boundaries and configuration

2. Code Flow Tracing

Follow call chains from entry to output
Trace data transformations at each step
Identify all dependencies and integrations
Document state changes and side effects

3. Architecture Analysis

Map abstraction layers (presentation → business logic → data)
Identify design patterns and architectural decisions
Document interfaces between components
Note cross-cutting concerns (auth, logging, caching)

Output Format

Entry Points

| Type | Location | Description | |------|----------|-------------| | API | src/api/users.py:45 | GET /users endpoint |

Execution Flow

1. [Entry] src/api/users.py:45 - get_users()
   ↓ validates request params
2. [Service] src/services/user_service.py:30 - list_users()
   ↓ applies business rules
3. [Repository] src/repositories/user_repo.py:55 - find_all()
   ↓ builds query
4. [Response] returns domain objects

Key Components

| Component | File | Responsibility | |-----------|------|----------------| | Service | src/services/x.py | Business logic | | Repository | src/repositories/x.py | Data access |

Dependencies

Internal: module → module relationships
External: libraries and frameworks

Observations

Strengths: what's well designed
Issues/Tech Debt: problems found
Essential Files: ordered list for further reading

Principles

LSP 도구를 기본으로 사용 (goToDefinition, findReferences, incomingCalls 등). 에러 시에만 Grep fallback
Always include file:line references
Trace complete flow, not just surface level
Identify patterns, not just code
Note both strengths and issues

What to Avoid

Using Read directly without understanding structure first
Reading entire files indiscriminately
LSP 시도 없이 바로 Grep/Glob부터 사용하는 것

Completion

실행 흐름 분석 리포트(Entry Points + Execution Flow + Key Components + Dependencies)가 file:line 참조와 함께 전달되면 완료.

Troubleshooting

LSP 도구가 에러 반환: 해당 파일 타입의 LSP 서버가 지원되지 않는 경우. Glob → Grep → Read로 fallback. "LSP 미지원 파일 타입 — Grep fallback" 1줄 고지. Codebase too large to trace completely: Start from the specific entry point the user cares about. Trace one flow end-to-end rather than mapping everything. Dynamic dispatch makes tracing difficult: Search for string patterns (route registrations, event handlers, DI bindings) that wire dynamic calls. Document as "dynamic dispatch — runtime-dependent" in the report.

Agent Skills: Code Explorer - Deep Codebase Analyst

Install this agent skill to your local

Skill Files