Multi-lens code review using agent teams for thorough, multi-perspective verification. Spawns reviewer, security, and architect agents as team members who independently review and then exchange findings via SendMessage for deeper coverage. Triggers on: "멀티렌즈 리뷰", "팀 리뷰", "team-review", "다관점 리뷰", or when a large/high-risk change needs comprehensive review from multiple perspectives.
Writes, modifies, diagnoses, and patches code following project conventions. Handles implementation, bug fixing, refactoring, and lint/format. Called by Orchestrator for all code change tasks, or directly by users. Triggers on: "코드 작성", "코드 수정", "code", or any direct request to write or modify code without a broader workflow.
Performs thorough deep research on any topic using web search, Context7 docs, and GitHub CLI. Saves structured results to docs/ directory. Use when the user asks to research, investigate, or gather comprehensive information on a topic. Triggers on: "조사해줘", "찾아줘", "알아봐", "리서치", "deepresearch", or any request for in-depth information gathering about a technology, library, framework, concept, or trend.
Synchronizes harness documentation (CLAUDE.md, harness.md) and install script (harness.sh) with the actual state of .claude/ subdirectories (skills, agents, hooks). Detects mismatches between source files and docs, then auto-updates affected sections. Triggers on: "문서 동기화", "docs sync", "sync-docs", or after harness file changes.
Analyzes current changes and creates a well-structured git commit. Use when the user wants to commit current changes with a proper message. Triggers on: "커밋해줘", "커밋", "commit", or any request to commit changes.
Performs evidence-based decision making by researching options, analyzing trade-offs, and providing clear recommendations with reasoning. Use when the user needs to choose between alternatives, evaluate options, or make informed decisions. Triggers on: "판단해줘", "결정해줘", "골라줘", "비교해줘", "추천해줘", "뭐가 나아", "어떤 걸 써야", "decide", or any request asking for comparison, recommendation, or decision support on technologies, tools, strategies, or approaches.
Parallel QA using agent teams for comprehensive multi-type testing. Spawns specialized QA agents that simultaneously run different test categories (unit, integration, lint, security scan) and consolidate results. Triggers on: "QA 스웜", "전체 QA", "qa-swarm", "병렬 테스트", "종합 테스트", or when comprehensive multi-category testing is needed for a project.
Scaffolds new skills and agents following established hoodcat-harness patterns. Reads existing files at runtime to ensure generated code matches current conventions. Triggers on: "스캐폴드", "스킬 생성", "에이전트 생성", "scaffold", or any request to create a new skill, agent, or skill+agent pair.