Initial commit

2025-11-29 18:19:59 +08:00
commit 9f60012661
14 changed files with 824 additions and 0 deletions
--- a/skills/scaffold-claude-feature/SKILL.md
+++ b/skills/scaffold-claude-feature/SKILL.md
@@ -0,0 +1,47 @@
+---
+name: scaffold-claude-feature
+description: Claude Code 확장 기능(skill, slash command, subagent, hook, output style)의 기본 구조를 생성합니다. 새로운 skill을 만들거나, command 템플릿이 필요하거나, hook 설정을 시작할 때 사용합니다.
+---
+
+# Claude Code Feature Scaffolder
+
+Quick scaffolder for Claude Code features with proper structure and best practices.
+
+## Feature Types
+
+1. **Slash Commands** - Reusable prompts (`.claude/commands/`)
+2. **Skills** - Model-invoked capabilities (`.claude/skills/`)
+3. **Subagents** - Specialized AI agents (`.claude/agents/`)
+4. **Hooks** - Event handlers (settings.json)
+5. **Output Styles** - Custom system prompts (`.claude/output-styles/`)
+
+## Workflow
+
+1. **Detect feature type** from user request
+2. **Ask location**: project (`.claude/`) or personal (`~/.claude/`)
+3. **Load reference**: See appropriate file in [references/](references/)
+4. **Generate feature** using [templates/](templates/)
+5. **Validate** naming and structure
+6. **Provide** testing guidance
+
+## Feature Detection
+
+- Keywords: "create", "generate", "add", "new"
+- Feature names: "slash command", "skill", "subagent", "agent", "hook", "output style"
+- Context: "Claude Code tool", "build tool", "custom feature"
+
+## Progressive Disclosure
+
+Load details only when needed:
+- Slash commands → [references/slash-commands.md](references/slash-commands.md)
+- Skills → [references/skills.md](references/skills.md)
+- Subagents → [references/subagents.md](references/subagents.md)
+- Hooks → [references/hooks.md](references/hooks.md)
+- Output styles → [references/output-styles.md](references/output-styles.md)
+
+## Quick Rules
+
+- Use lowercase with hyphens for names
+- Project features for teams, personal for individual use
+- Keep examples simple and minimal
+- Test after creation
--- a/skills/scaffold-claude-feature/references/hooks.md
+++ b/skills/scaffold-claude-feature/references/hooks.md
@@ -0,0 +1,167 @@
+# Hooks Reference
+
+Quick reference for creating event hooks. For complete documentation, see @docs-claude/hooks-reference.md
+
+## Structure
+
+Add to settings.json:
+```json
+{
+  "hooks": {
+    "EventName": [
+      {
+        "matcher": "ToolPattern",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "your-command"
+          }
+        ]
+      }
+    ]
+  }
+}
+```
+
+## Location
+
+- Project: `.claude/settings.json`
+- Personal: `~/.claude/settings.json`
+- Local: `.claude/settings.local.json` (gitignored)
+
+## Hook Types
+
+- `"type": "command"` - Execute bash command
+- `"type": "prompt"` - LLM-based evaluation
+
+## Common Events
+
+- `PreToolUse` - Before tool runs (can block)
+- `PostToolUse` - After tool completes
+- `UserPromptSubmit` - Before processing user prompt
+- `Stop` - When Claude finishes
+- `SubagentStop` - When subagent finishes
+- `Notification` - On notifications
+- `SessionStart` - Session begins
+- `SessionEnd` - Session ends
+
+## Matchers
+
+**For PreToolUse/PostToolUse:**
+- `"Write"` - Exact match
+- `"Write|Edit"` - Multiple tools
+- `"Notebook.*"` - Regex pattern
+- `"*"` - All tools
+
+**For other events:** Omit matcher
+
+## Exit Codes (Command Type)
+
+- `0` - Success
+- `2` - Block action, show stderr to Claude
+- Other - Error, execution continues
+
+## Simple Examples
+
+**Format on write:**
+```json
+{
+  "hooks": {
+    "PostToolUse": [
+      {
+        "matcher": "Write|Edit",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "prettier --write $file"
+          }
+        ]
+      }
+    ]
+  }
+}
+```
+
+**Prevent stopping:**
+```json
+{
+  "hooks": {
+    "Stop": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "$CLAUDE_PROJECT_DIR/.claude/hooks/check-todos.sh"
+          }
+        ]
+      }
+    ]
+  }
+}
+```
+
+**Validate bash commands:**
+```json
+{
+  "hooks": {
+    "PreToolUse": [
+      {
+        "matcher": "Bash",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "python $CLAUDE_PROJECT_DIR/.claude/hooks/validate-bash.py"
+          }
+        ]
+      }
+    ]
+  }
+}
+```
+
+## Hook Input
+
+Hooks receive JSON via stdin:
+```json
+{
+  "session_id": "abc123",
+  "hook_event_name": "PreToolUse",
+  "tool_name": "Write",
+  "tool_input": { "file_path": "/path/to/file" }
+}
+```
+
+## Environment Variables
+
+- `$CLAUDE_PROJECT_DIR` - Project root path
+- `$CLAUDE_ENV_FILE` - SessionStart only, for env vars
+
+## Tips
+
+- Use `$CLAUDE_PROJECT_DIR` for project scripts
+- Test hooks with `claude --debug`
+- Exit code 2 blocks and sends stderr to Claude
+- Keep hooks fast (60s timeout default)
+- For complete API, see @docs-claude/hooks-reference.md
+
+## Prompt-Based Hooks
+
+Use LLM for intelligent decisions:
+```json
+{
+  "hooks": {
+    "Stop": [
+      {
+        "hooks": [
+          {
+            "type": "prompt",
+            "prompt": "Check if all tasks complete: $ARGUMENTS"
+          }
+        ]
+      }
+    ]
+  }
+}
+```
+
+LLM responds with: `{"decision": "approve"|"block", "reason": "..."}`
--- a/skills/scaffold-claude-feature/references/output-styles.md
+++ b/skills/scaffold-claude-feature/references/output-styles.md
@@ -0,0 +1,119 @@
+# Output Styles Reference
+
+Quick reference for creating custom output styles.
+
+## Structure
+
+Markdown file with frontmatter:
+```markdown
+---
+name: My Style
+description: Brief description shown in UI
+keep-coding-instructions: false
+---
+
+# Custom Style Instructions
+
+Your custom system prompt additions here.
+Define how Claude should behave in this style.
+```
+
+## Location
+
+- Project: `.claude/output-styles/style-name.md`
+- Personal: `~/.claude/output-styles/style-name.md`
+
+## Key Fields
+
+- `name` - Display name (optional, defaults to filename)
+- `description` - Shown in `/output-style` menu
+- `keep-coding-instructions` - Keep Claude's coding instructions (default: false)
+
+## Activation
+
+- `/output-style` - Open menu to select
+- `/output-style style-name` - Switch directly
+- Saved in `.claude/settings.local.json`
+
+## How It Works
+
+Output styles modify the system prompt:
+- Removes efficiency instructions (concise output)
+- Removes coding instructions (unless keep-coding-instructions: true)
+- Adds your custom instructions
+
+## Simple Examples
+
+**Teacher mode:**
+```markdown
+---
+name: Teacher
+description: Educational mode with explanations
+keep-coding-instructions: true
+---
+
+# Teacher Mode
+
+You help users learn while coding.
+
+Behavior:
+- Explain your reasoning
+- Share insights between tasks
+- Ask questions to check understanding
+- Encourage hands-on practice
+```
+
+**Formal mode:**
+```markdown
+---
+name: Formal
+description: Professional, detailed responses
+---
+
+# Formal Mode
+
+You provide formal, professional assistance.
+
+Style:
+- Use formal language
+- Provide detailed explanations
+- Structure responses clearly
+- Include relevant references
+```
+
+**Quick mode:**
+```markdown
+---
+name: Quick
+description: Minimal, direct responses
+keep-coding-instructions: true
+---
+
+# Quick Mode
+
+Provide minimal, direct responses.
+
+- Short answers
+- Code only when needed
+- No extra explanations
+```
+
+## Tips
+
+- Set `keep-coding-instructions: true` for coding tasks
+- Clear behavior guidelines work best
+- Test with different request types
+- Use for non-coding tasks (research, writing, etc.)
+
+## Built-in Styles
+
+- **Default** - Standard software engineering
+- **Explanatory** - Adds educational insights
+- **Learning** - Collaborative learn-by-doing
+
+## Comparison
+
+- **Output Styles** - Change system prompt, affect all responses
+- **CLAUDE.md** - Add context, doesn't change system prompt
+- **Subagents** - Separate tasks, own context
+- **Slash Commands** - Reusable prompts
--- a/skills/scaffold-claude-feature/references/skills.md
+++ b/skills/scaffold-claude-feature/references/skills.md
@@ -0,0 +1,109 @@
+# Skills Reference
+
+Quick reference for creating agent skills.
+
+## Structure
+
+```
+my-skill/
+├── SKILL.md (required)
+├── reference.md (optional)
+├── examples.md (optional)
+├── scripts/
+│   └── helper.py
+└── templates/
+    └── template.txt
+```
+
+## Location
+
+- Project: `.claude/skills/skill-name/`
+- Personal: `~/.claude/skills/skill-name/`
+
+## SKILL.md Format
+
+```markdown
+---
+name: my-skill-name
+description: What it does and when to use it. Include trigger keywords.
+allowed-tools: Read, Write, Bash
+---
+
+# My Skill
+
+## Instructions
+
+1. Clear step-by-step guidance
+2. What to do and how
+
+## Examples
+
+Simple, minimal examples
+```
+
+## Description Tips
+
+Include WHEN to use:
+- ✅ "Extract PDF text. Use when working with PDF files."
+- ❌ "Extract PDF text."
+
+## Tool Access
+
+- Omit `allowed-tools` → inherit all tools
+- Specify tools → restrict access (e.g., read-only: `Read, Grep, Glob`)
+
+## Progressive Disclosure
+
+Reference other files:
+```markdown
+For advanced usage, see [reference.md](reference.md)
+Run helper: `python scripts/helper.py`
+```
+
+Claude loads files only when needed.
+
+## Simple Examples
+
+**Code reviewer:**
+```markdown
+---
+name: code-reviewer
+description: Review code quality. Use after writing or modifying code.
+allowed-tools: Read, Grep, Glob
+---
+
+# Code Reviewer
+
+## Checklist
+- Readable code
+- Error handling
+- No duplicated code
+- Security issues
+
+Review and provide specific feedback.
+```
+
+**Commit helper:**
+```markdown
+---
+name: commit-helper
+description: Generate commit messages. Use when staging commits.
+allowed-tools: Bash(git diff:*), Bash(git status:*)
+---
+
+# Commit Helper
+
+## Instructions
+1. Run `git diff --staged`
+2. Generate clear commit message:
+   - Summary under 50 chars
+   - Present tense
+   - Explain what and why
+```
+
+## Tips
+
+- One skill = one focused capability
+- Clear, specific descriptions
+- Simple instructions
+- Test with relevant requests
--- a/skills/scaffold-claude-feature/references/slash-commands.md
+++ b/skills/scaffold-claude-feature/references/slash-commands.md
@@ -0,0 +1,85 @@
+# Slash Commands Reference
+
+Quick reference for creating custom slash commands.
+
+## Structure
+
+```markdown
+---
+description: Brief description
+argument-hint: [arg1] [arg2]
+allowed-tools: Tool1, Tool2
+---
+
+Your command prompt here with $ARGUMENTS or $1, $2, etc.
+```
+
+## Location
+
+- Project: `.claude/commands/filename.md` → `/filename`
+- Personal: `~/.claude/commands/filename.md` → `/filename`
+- Subdirectories: `.claude/commands/subdir/filename.md` → `/filename` (subdir shown in description)
+
+## Arguments
+
+- `$ARGUMENTS` - All arguments as single string
+- `$1, $2, $3` - Individual arguments
+
+## File References
+
+Use `@` prefix to reference files:
+```
+Review @src/app.js for bugs
+Compare @old.js with @new.js
+```
+
+## Bash Execution
+
+Prefix with `!` to execute bash:
+```markdown
+---
+allowed-tools: Bash(git status:*), Bash(git diff:*)
+---
+
+Current status: !`git status`
+Recent changes: !`git diff HEAD`
+```
+
+## Simple Examples
+
+**Commit helper:**
+```markdown
+---
+description: Create git commit
+allowed-tools: Bash(git add:*), Bash(git commit:*)
+---
+
+Create a commit for these changes: !`git diff`
+```
+
+**Review PR:**
+```markdown
+---
+description: Review pull request
+argument-hint: [pr-number]
+---
+
+Review PR #$1 focusing on code quality and security.
+```
+
+**Fix issue:**
+```markdown
+---
+description: Fix GitHub issue
+argument-hint: [issue-number]
+---
+
+Fix issue #$ARGUMENTS following our coding standards.
+```
+
+## Tips
+
+- Keep descriptions short and clear
+- Use `allowed-tools` to pre-approve tools
+- Simple prompts work best
+- Test with different arguments
--- a/skills/scaffold-claude-feature/references/subagents.md
+++ b/skills/scaffold-claude-feature/references/subagents.md
@@ -0,0 +1,115 @@
+# Subagents Reference
+
+Quick reference for creating specialized subagents.
+
+## Structure
+
+Single Markdown file:
+```markdown
+---
+name: my-agent
+description: What it does. Use proactively when [condition].
+tools: Read, Write, Bash
+model: sonnet
+---
+
+System prompt for this subagent.
+Detailed role and approach instructions.
+```
+
+## Location
+
+- Project: `.claude/agents/agent-name.md`
+- Personal: `~/.claude/agents/agent-name.md`
+
+## Key Fields
+
+- `name` - lowercase-with-hyphens
+- `description` - Include "use proactively" for automatic invocation
+- `tools` - Optional; omit to inherit all tools
+- `model` - Optional: `sonnet`, `opus`, `haiku`, or `inherit`
+
+## Model Selection
+
+- `sonnet` (default) - Balanced capability
+- `opus` - Maximum capability
+- `haiku` - Fast, cost-effective
+- `inherit` - Use main conversation's model
+
+## Invocation
+
+**Automatic:** Claude delegates based on description
+**Explicit:** User mentions agent by name
+
+## Simple Examples
+
+**Code reviewer:**
+```markdown
+---
+name: code-reviewer
+description: Expert code reviewer. Use proactively after writing code.
+tools: Read, Grep, Glob, Bash(git diff:*)
+model: inherit
+---
+
+You are a senior code reviewer.
+
+When invoked:
+1. Run git diff to see changes
+2. Review for quality and security
+3. Provide specific, actionable feedback
+
+Focus on:
+- Readable code
+- Error handling
+- Security issues
+- Performance
+```
+
+**Debugger:**
+```markdown
+---
+name: debugger
+description: Debug errors and test failures. Use proactively when issues occur.
+tools: Read, Edit, Bash, Grep
+---
+
+You are an expert debugger.
+
+Process:
+1. Capture error message
+2. Identify cause
+3. Implement minimal fix
+4. Verify solution
+
+Provide root cause and fix.
+```
+
+**Test runner:**
+```markdown
+---
+name: test-runner
+description: Run tests and fix failures. Use proactively after code changes.
+---
+
+You run tests and fix failures.
+
+Steps:
+1. Run appropriate tests
+2. Analyze failures
+3. Fix issues
+4. Re-run tests
+```
+
+## Tips
+
+- Single, clear responsibility per agent
+- Include "use proactively" for automatic use
+- Limit tools to what's needed
+- Test with relevant requests
+- Use `/agents` command to manage
+
+## When to Use
+
+- **Subagents** - Separate context, specialized tasks
+- **Skills** - Extend capabilities, loaded in main context
--- a/skills/scaffold-claude-feature/templates/hooks.template.json
+++ b/skills/scaffold-claude-feature/templates/hooks.template.json
@@ -0,0 +1,16 @@
+{
+  "hooks": {
+    "[EventName]": [
+      {
+        "matcher": "[ToolPattern or omit for events like Stop]",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "[bash-command or $CLAUDE_PROJECT_DIR/script.sh]",
+            "timeout": 60
+          }
+        ]
+      }
+    ]
+  }
+}
--- a/skills/scaffold-claude-feature/templates/output-style.template.md
+++ b/skills/scaffold-claude-feature/templates/output-style.template.md
@@ -0,0 +1,16 @@
+---
+name: [Style Name]
+description: [Brief description shown in /output-style menu]
+keep-coding-instructions: [true or false]
+---
+
+# [Style Name]
+
+[Your custom system prompt instructions here]
+
+Define how Claude should behave in this style:
+- [Behavior 1]
+- [Behavior 2]
+- [Behavior 3]
+
+[Additional guidelines and approach]
--- a/skills/scaffold-claude-feature/templates/skill.template.md
+++ b/skills/scaffold-claude-feature/templates/skill.template.md
@@ -0,0 +1,20 @@
+---
+name: [skill-name-lowercase-with-hyphens]
+description: [What it does and when to use it. Include trigger keywords.]
+allowed-tools: [Optional: Read, Write, Bash - omit to inherit all tools]
+---
+
+# [Skill Name]
+
+## Instructions
+
+1. [Step-by-step guidance]
+2. [What to do and how]
+
+## Examples
+
+[Simple, minimal examples]
+
+## Tips
+
+[Best practices or common patterns]
--- a/skills/scaffold-claude-feature/templates/slash-command.template.md
+++ b/skills/scaffold-claude-feature/templates/slash-command.template.md
@@ -0,0 +1,11 @@
+---
+description: [Brief description of what this command does]
+argument-hint: [arg1] [arg2]
+allowed-tools: [Tool1, Tool2]
+---
+
+[Your command prompt here]
+
+Use $ARGUMENTS for all args, or $1, $2 for individual args.
+Use @filepath to reference files.
+Use !`command` to execute bash (requires allowed-tools).
--- a/skills/scaffold-claude-feature/templates/subagent.template.md
+++ b/skills/scaffold-claude-feature/templates/subagent.template.md
@@ -0,0 +1,20 @@
+---
+name: [agent-name-lowercase-with-hyphens]
+description: [What it does. Use proactively when [condition].]
+tools: [Optional: Read, Write, Bash - omit to inherit all tools]
+model: [Optional: sonnet, opus, haiku, or inherit]
+---
+
+You are [role description].
+
+When invoked:
+1. [Step 1]
+2. [Step 2]
+3. [Step 3]
+
+Focus on:
+- [Key aspect 1]
+- [Key aspect 2]
+- [Key aspect 3]
+
+[Additional guidance and approach]