faion-claude-code（faion Claude 代码）

Entry point: /faion-net — invoke this skill for automatic routing to the appropriate domain.

Claude Code Configuration Skill

Communication: User's language. Config/code: English.

Purpose

Orchestrate Claude Code customization and configuration:

• Create/edit skills, agents, commands, hooks
• Configure settings, permissions, IDE integrations
• Setup MCP servers and plugins
• Enforce naming conventions

Context Discovery

Auto-Investigation

Discovery Questions

questions:
  - question: "What are you configuring?"
    options:
      - label: "Creating skill"
        description: "Use skills.md"
      - label: "Creating agent"
        description: "Use agents.md"
      - label: "Adding command"
        description: "Use commands.md"
      - label: "Setting up hooks"
        description: "Use hooks.md"
      - label: "Installing MCP server"
        description: "Use mcp-servers.md"

  - question: "Is this global or project-specific?"
    options:
      - label: "Global (all projects)"
        description: "Use faion-* prefix, no gitignore"
      - label: "Project-specific"
        description: "Use {project}-* prefix, add to gitignore"

  - question: "Do you have Claude Code experience?"
    options:
      - label: "Expert user"
        description: "Jump to specific reference"
      - label: "Some experience"
        description: "Use quick reference + examples"
      - label: "First time setup"
        description: "Start with skills.md basics"

Quick Decision Tree

Configuration Locations:

References

Detailed technical context for specialized areas:

Total: ~1,960 lines of technical reference

Routing

User Request → Detect Type → Load Reference

Naming Conventions

Global (Faion Network)

For shared/reusable components in faion-network:

Note: Skills use role/process naming without -skill suffix. Name is self-explanatory.

Project-Specific (Local)

For project-specific components that should NOT be committed to faion-network:

Setup for project-specific components:

1. Add to.gitignore (same level as.claude/):

echo ".claude/skills/{project}-*/" >> .gitignore
echo ".claude/agents/{project}-*" >> .gitignore
echo ".claude/commands/{project}-*" >> .gitignore
echo ".claude/scripts/hooks/{project}-*" >> .gitignore

1. Add attribution footer:

---
*Created with [faion.net](https://faion.net) framework*

Rules Summary

Directory Structure

~/.claude/
├── settings.json           # Global settings
├── CLAUDE.md              # Global instructions
├── agents/
│   └── faion-*-agent.md   # Agent definitions
├── skills/
│   └── faion-*-skill/
│       ├── SKILL.md       # Skill index
│       └──     # Detailed content
├── commands/
│   └── *.md               # Slash commands
└── scripts/
    └── hooks/             # Hook scripts

Settings Configuration

settings.json Structure

{
  "permissions": {
    "allow": ["Read", "Glob", "Grep"],
    "deny": ["Bash(rm -rf:*)"]
  },
  "hooks": {
    "PreToolUse": [...],
    "PostToolUse": [...],
    "SessionStart": [...]
  },
  "env": {
    "CUSTOM_VAR": "value"
  },
  "mcpServers": {
    "server-name": {
      "command": "npx",
      "args": ["-y", "package-name"],
      "env": { "API_KEY": "..." }
    }
  }
}

Permission Modes

Configuration Locations

Quick Reference

Create Skill

mkdir -p ~/.claude/skills/faion-my-skill
# Write SKILL.md with frontmatter

Create Agent

# Write ~/.claude/agents/faion-my-agent.md
# Include frontmatter: name, description, tools, model

Create Command

# Write ~/.claude/commands/my-cmd.md
# Include frontmatter: description, argument-hint, allowed-tools

Create Hook

# Write ~/.claude/scripts/hooks/faion-pre-bash-my-hook.py
# Configure in settings.json hooks section

Install MCP Server

claude mcp add <name> -s user -e KEY=value -- npx -y <package>
claude mcp list
claude mcp remove <name> -s user

IDE Integrations

VS Code

// .vscode/settings.json
{
  "claude.enabled": true,
  "claude.autoComplete": true
}

JetBrains

Install Claude plugin from marketplace.

Best Practices

Skills:

• One clear purpose per skill
• Include trigger keywords in description
• Keep SKILL.md under 300 lines (use)
• Third-person descriptions only

Agents:

• Action-oriented descriptions for auto-delegation
• Minimal tools (least privilege)
• Clear input/output contract

Commands:

• Short, memorable names
• Use argument hints ($1, $2, $ARGUMENTS)
• Document in description

Hooks:

• Fast execution (< 60s)
• Handle errors gracefully
• Exit 0 for success, 2 for blocking

Error Handling

Documentation

• Claude Code Docs
• Skills
• Sub-agents
• Hooks
• MCP Servers

*faion-claude-code-skill v2.0.0* *Claude Code configuration orchestrator with *