메뉴
Configure Claude Code hooks for automatic formatting, safety guards, compaction recovery, and session management. Use when the user says "set up hooks", "auto-format", "block dangerous commands", "configure automation", "safety rules", or wants deterministic automation.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | hooks-setup |
| description | Configure Claude Code hooks for automatic formatting, safety guards, compaction recovery, and session management. Use when the user says "set up hooks", "auto-format", "block dangerous commands", "configure automation", "safety rules", or wants deterministic automation. |
| argument-hint | [auto-format|safety|compaction-recovery|session|all] |
| disable-model-invocation | true |
| allowed-tools | Read, Write, Edit, Glob |
Interactive setup of Claude Code hooks — deterministic automation that runs without AI reasoning tokens.
auto-format — Format code after every editPostToolUse on Edit/Write. Detects language from file extension, runs appropriate formatter.
ruff formatprettier --writegofmt -wsafety — Block destructive commandsPreToolUse on Bash. Blocks: rm -rf /, rm -rf ~, git push --force main, DROP TABLE, DROP DATABASE, --break-system-packages.
compaction-recovery — Re-inject context after compactionSessionStart on compact. Creates .claude/context-essentials.md with critical project context, echoes it after compaction so Claude stays oriented.
session — Session start/end management/wrap-up.claude/settings.json (if exists)For each category, generate the hook JSON. Show the user exactly what will be added.
.claude/settings.jsonWalk user through creating .claude/context-essentials.md:
This file gets echoed into context after every compaction, keeping Claude oriented.
Browse, install, and uninstall skills and agents from the available library. Check for new ones online. Use when the user says "what skills are available", "install skill", "add skill", "remove skill", "skill catalog", "list available", "check for updates", or wants to set up a project with specific capabilities.
Install Claude Code expert toolkit into an existing project. Use when the user says "add to my project", "install into existing", "set up existing project", "integrate", "add skills to this project", or wants to enhance an already-started codebase with Claude Code best practices.
Set up lat.md knowledge graph for a project. lat.md replaces flat AGENTS.md/CLAUDE.md with an interconnected, validated graph of markdown docs linked to source code. Use when user says "set up lat.md", "install lat.md", "add knowledge graph", "lat md", "agent lattice", or wants structured, validated project documentation that stays in sync with code. Also triggered via "lat-md-setup checklist" to return just the recommendation checklist.
Create a fully configured new project with Claude Code best practices, skills, and agents. Use when the user says "new project", "start a project", "create a project", "set up a new", or "bootstrap a project".
Transpile source code into Claude Code markdown skills. Takes Python, JavaScript, TypeScript, C++, Rust, Go, or any terminal-based program and converts it into an LLM-executable markdown skill. Analyzes code structure, fuzzes runtime behavior, classifies operations, and generates a complete skill directory with SKILL.md, bundled scripts, and reference docs. Use when user says "compile to skill", "convert this code to a skill", "make a skill from this script", "turn this into a skill", "transpile to skill", or wants to translate any traditional program into Claude Code's markdown skill format. Also use when user pastes code and asks to make it work as a Claude skill.
Run commands requiring sudo in a visible terminal window where the user can enter their password interactively. Use when a command needs root privileges, elevated permissions, "sudo", "install system package", "apt install", "dnf install", "pacman -S", or any operation requiring admin access.