skills

Diagnose macOS RAM by memory pressure rather than raw usage — runs one script that reports the pressure verdict, swap, the usage-vs-cache breakdown, top consumers, and an optional process-group total. Use when the user asks about memory or RAM, "why is my RAM so high", "is 36GB/48GB enough", whether to worry about usage, 메모리 사용량, 램 사용량 높은데, 메모리 부족한가, 맥 메모리 확인.

Compresses a long output, report, or analysis into a one-page executive (CEO/CTO) brief — conclusion first, key points only, technical detail translated into business impact. Use when the user wants the gist of something too long for a decision-maker, with phrases like 내용이 너무 많아서 읽기 힘들다, 한눈에 안 들어온다, 다 읽기 귀찮으니 요약해줘, CEO/CTO에게 보고하듯, 한 페이지로 정리해줘, 핵심만 압축해서 알려줘, 정리해줘, 요약해줘.

Unified GitHub issue workflow via the gh CLI — fetch an issue (body, labels, comments, image/video attachments) into the session, create a new issue from a template, or update an existing one (body, labels, state, comments). Handles the 404 that plain curl/WebFetch hits on GitHub attachment URLs by downloading with an auth token, and extracts viewable frames from video attachments. Use whenever the user mentions GitHub issues — 이슈 가져와, 이슈 조회, 이슈 등록, 이슈 생성, 이슈 만들어줘, 이슈 업데이트, 이슈 수정, 이슈 코멘트, 이슈 작업, fetch issue N, work on issue N, create an issue, update issue N, file a bug, /github-issue.

Writes single-scenario Flow documents under flows/<topic>/ — use case context, one Mermaid diagram (sequence or flowchart), step-by-step branches, and source file references — so how a service behaves end-to-end can be grasped at a glance. Use when user asks for a sequence diagram, user flow, use case doc, or mentions 시퀀스 다이어그램, 유저 플로우, 플로우 그려줘, 동작 흐름, 플로우 작성. Also trigger on intent like 어떻게 동작하는지 한눈에 보고 싶다, 동작 순서 정리해줘 — but not on simple code questions about a single function.

Creates and manages an agent team (planner + implementer teammates) for the current worktree session. Use when user says "create team", "팀 생성", "팀 만들어줘", "agent team", or wants a planner+implementer spec-driven workflow in a worktree session. Also covers team operations — cleanup ("팀 정리", end teammates) and expand ("팀원 추가", add a role) — folded from the former agent-team commands.

Splits current project work into parallel-safe task groups with worktree branch names and structured starting prompts. Use only on manual /split-work invocation — no automatic trigger.

Interactive headed-browser walkthrough with playwright-cli — attaches to the user's browser and advances one step per user confirmation, with explicit keyword gates for irreversible actions like final submit or payment. Use for iframe-heavy Korean sites (홈택스, 정부24, 은행, 쇼핑몰) done together with the user, or when the user says 브라우저 보면서 같이 진행, headed 모드로 같이, 단계별로 진행해줘, 홈택스 같이, 정부24 같이, 한 스텝씩, 한 단계씩 진행, walkthrough 모드, or step-by-step browser.

Runs app test scenarios through Computer Use MCP and produces a structured UI/UX feedback report with screenshot evidence. Use when the user says computer use 테스트, 앱 테스트, UI 테스트, cu test, computer use test, or 앱 QA.

Generates a distinctive standalone HTML preview from a saved UI brainstorm idea. Use when the user wants to visualize a brainstorm idea as a concrete mockup, prototype, preview, landing page, dashboard, or product screen with strong design quality.

Talk to and control the Hermes companion runtime — chat, run async tasks, check status, manage cron jobs, configure the connection. Use for /hermes, "hermes chat", "hermes run", "hermes status", "hermes jobs", "hermes setup", or 헤르메스 채팅/실행/상태/잡/설정.

Brainstorms grounded feature ideas from the current codebase and saves them as wireframed notes in `brain-storm/` — the ideation step before writing specs. Use when the user asks to brainstorm ideas, generate feature ideas, explore improvements, review future opportunities, or clean up outdated brainstorm notes.

Maintains an LLM-powered 3-layer personal wiki (raw → wiki → schema) from raw sources. Use when user mentions wiki init, wiki ingest, wiki query, wiki lint, wiki update, 위키 초기화, 위키 추가, 위키 질문, 위키 검사, 위키 업데이트, or wants to build a knowledge base from sources.

Locates a previous Claude Code or Codex CLI session's JSONL transcript and prints the last N turns plus metadata so work can continue. Use when the user says "resume", "이어가자", "이전 세션 내용 보여줘", "what was I working on", or provides a session UUID.

Implements a task as production code plus tests, auto-detecting the test framework (jest, vitest, pytest, go test, cargo test) and following existing project patterns. Task information comes from an argument (description or task file path) or the current conversation. Use when user asks to implement a task, build a feature with tests, or says "implement", "구현", "구현해줘", "테스트와 함께 구현". Proactively trigger whenever the user wants to turn a task or requirement into working code with tests.

Runs a branch-safe finish workflow that lints, tests, commits, pushes, creates a pull request, and cleans merged worktrees. Use when the user asks to finish work, test and commit changes, push a branch, open a PR, or clean completed worktrees.

Writes and manages spec files in specs/ with mandatory duplicate search, conflict detection, and a structured final report. Use when user asks to create a spec, update a spec, write a spec, or mentions 스펙 생성, 스펙 업데이트, 스펙 작성, 스펙 만들어줘. Proactively trigger whenever the request involves specification documents, even if the user never says spec.

Decomposes specs into persistent task files under tasks/ with a dependency graph, parallel lane suggestions, and dynamically derived progress. Use when user mentions task 분리, task 생성, task 쪼개기, spec을 task로, decompose, writing-tasks, 병렬 작업 편성, 의존성 그래프. Proactively trigger when a specs/ directory has unmapped specs or when user wants to track long-running work as persistent task files.

Install macOS TTS + dialog notifications for Claude Code Stop and Notification events — a voice cue and a clickable dialog when Claude finishes a task or needs permission/input. Use when the user wants desktop or voice notifications, "set up notifications", "작업 끝나면 알려줘", "소리로 알려줘", or to wire up the Stop/Notification hooks.

Comprehensive code quality review focusing on DRY (Don't Repeat Yourself), KISS (Keep It Simple, Stupid), and Clean Code (Easy to Read) principles. Use when reviewing code changes, analyzing code quality, or assessing code maintainability for any feature, bug fix, or refactoring task.

Code style principle-based review - checks SRP (Single Responsibility Principle), DRY (Don't Repeat Yourself), Simplicity First, YAGNI (You Aren't Gonna Need It), and Type Safety. Also evaluates code structure and naming conventions. Automatically used when code review is needed.

Reviews React components and provides improvement suggestions. Use when you need code review, component review, structure improvement, refactoring, or component separation.

Manage git worktrees for parallel branch work. PROACTIVELY USE when user mentions working on a PR, new feature, or new task - ask if they want to create a worktree BEFORE starting implementation.

Symlink 기반 로컬 플러그인 테스트 - 설치/해제 없이 개발 디렉토리를 직접 연결

Generate basic structure for Claude Code extension features (skill, slash command, subagent, hook, output style). Use when creating a new skill, needing a command template, or starting hook configuration.

Generate HTML file to view work session. Use when user asks to view content as HTML (e.g., 'view as HTML', 'export to HTML', 'create HTML file', 'save as HTML').

Split uncommitted working tree changes into logical, ordered commits by analyzing Claude Code session history. Use when user asks to split commits, organize commits from session, auto-commit by feature, or says "커밋 나눠줘", "커밋 분리", "세션 기반 커밋".

Manage specification files for spec-driven development. Use when creating, updating, or deleting spec files. Supports feat, fix, docs, refactor, and chore scopes.

Extract Claude Code work history and update daily notes. Use when user asks to update daily file with today's work, sync work history, record activities, or generate daily summary from Claude Code history.