moai-cc-settings
Configuring Claude Code settings.json & Security. Set up permissions (allow/deny), permission modes, environment variables, tool restrictions. Use when securing Claude Code, restricting tool access, or optimizing session settings.
When & Why to Use This Skill
This Claude skill provides a comprehensive framework for configuring Claude Code's settings.json to ensure maximum security and operational control. It focuses on establishing robust permission systems, managing environment variables securely, and implementing tool restrictions to prevent unauthorized or dangerous actions by the AI agent.
Use Cases
- Securing development environments by defining granular allow/deny lists for file access and bash command execution.
- Implementing 'ask-before-act' workflows to maintain human-in-the-loop oversight for critical system modifications.
- Preventing accidental data leaks by restricting the agent's ability to read sensitive files like .env, .ssh, or system secrets.
- Standardizing project-specific configurations across teams to ensure consistent safety protocols and tool availability.
- Configuring secure environment variable patterns to handle API keys and credentials without hardcoding them into the codebase.
| name | moai-cc-settings |
|---|---|
| description | "Configuring Claude Code settings.json & Security. Set up permissions (allow/deny), permission modes, environment variables, tool restrictions. Use when securing Claude Code, restricting tool access, or optimizing session settings." |
| allowed-tools | "Read, Write, Edit, Bash" |
Skill Metadata
| Field | Value |
|---|---|
| Version | 1.0.0 |
| Tier | Ops |
| Auto-load | When configuring security & permissions |
What It Does
settings.json 설정 및 보안 구성을 위한 전체 가이드를 제공합니다. Permissions (allow/deny), permission modes, environment variables, tool restrictions 설정 방법을 다룹니다.
When to Use
- 새 프로젝트의 settings.json을 설정할 때
- Tool access를 제한하거나 보안을 강화할 때
- Environment variables를 구성할 때
- Permission mode (ask/allow/deny)를 변경할 때
Configuring Claude Code settings.json
settings.json centralizes all Claude Code configuration: permissions, tool access, environment variables, and session behavior.
Location: .claude/settings.json
Complete Configuration Template
{
"permissions": {
"allowedTools": [
"Read(**/*.{js,ts,json,md})",
"Edit(**/*.{js,ts})",
"Glob(**/*)",
"Grep(**/*)",
"Bash(git:*)",
"Bash(npm:*)",
"Bash(npm run:*)",
"Bash(pytest:*)",
"Bash(python:*)"
],
"deniedTools": [
"Read(./.env)",
"Read(./.env.*)",
"Read(./secrets/**)",
"Bash(rm -rf:*)",
"Bash(sudo:*)",
"Bash(curl:*)"
]
},
"permissionMode": "ask",
"spinnerTipsEnabled": true,
"disableAllHooks": false,
"env": {
"ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
"GITHUB_TOKEN": "${GITHUB_TOKEN}",
"CLAUDE_CODE_ENABLE_TELEMETRY": "1"
},
"hooks": {
"PreToolUse": [
{
"matcher": "Bash",
"hooks": [
{
"type": "command",
"command": "bash ~/.claude/hooks/pre-bash-check.sh"
}
]
}
],
"PostToolUse": [
{
"matcher": "Edit",
"hooks": [
{
"type": "command",
"command": "bash ~/.claude/hooks/post-edit-format.sh"
}
]
}
],
"SessionStart": [
{
"matcher": "*",
"hooks": [
{
"type": "command",
"command": "bash ~/.claude/hooks/session-init.sh"
}
]
}
]
},
"statusLine": {
"enabled": true,
"type": "command",
"command": "~/.claude/statusline.sh"
},
"mcpServers": {
"github": {
"command": "npx",
"args": ["-y", "@anthropic-ai/mcp-server-github"],
"oauth": {
"clientId": "${GITHUB_CLIENT_ID}",
"clientSecret": "${GITHUB_CLIENT_SECRET}",
"scopes": ["repo", "issues"]
}
},
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "${CLAUDE_PROJECT_DIR}/.moai", "${CLAUDE_PROJECT_DIR}/src"]
}
},
"extraKnownMarketplaces": [
{
"name": "company-plugins",
"url": "https://github.com/your-org/claude-plugins"
}
]
}
Permission Modes
| Mode | Behavior | Use Case |
|---|---|---|
| allow | Execute all allowed tools without asking | Trusted environments |
| ask | Ask before executing each tool | Development (safer) |
| deny | Deny all tools except whitelisted | Restrictive (default) |
{
"permissionMode": "ask"
}
Tool Permission Patterns
Restrictive (Recommended for teams)
{
"allowedTools": [
"Read(src/**)",
"Edit(src/**/*.ts)",
"Bash(npm run test:*)",
"Glob(src/**)"
],
"deniedTools": [
"Bash(rm:*)",
"Bash(sudo:*)",
"Read(.env)"
]
}
Permissive (Local development only)
{
"allowedTools": [
"Read",
"Write",
"Edit",
"Bash(git:*)",
"Bash(npm:*)",
"Bash(python:*)",
"Glob",
"Grep"
]
}
Environment Variables Pattern
{
"env": {
"ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
"GITHUB_TOKEN": "${GITHUB_TOKEN}",
"BRAVE_SEARCH_API_KEY": "${BRAVE_SEARCH_API_KEY}",
"NODE_ENV": "development"
}
}
Security rule: Never hardcode secrets; always use ${VAR_NAME} syntax.
Dangerous Tools to Deny
{
"deniedTools": [
"Bash(rm -rf:*)", // Recursive delete
"Bash(sudo:*)", // Privilege escalation
"Bash(curl.*|.*bash)", // Code injection
"Read(.env)", // Secrets
"Read(.ssh/**)", // SSH keys
"Read(/etc/shadow)", // System secrets
"Edit(/etc/**)", // System files
]
}
Permission Validation
# Check current permissions
cat .claude/settings.json | jq '.permissions'
# Validate JSON syntax
jq . .claude/settings.json
# List allowed tools
jq '.permissions.allowedTools[]' .claude/settings.json
Spinner Tips Configuration
{
"spinnerTipsEnabled": true
}
Custom tips can be added for better UX during long operations.
Best Practices
✅ DO:
- Use
askmode for teams - Explicitly whitelist paths
- Environment variables for all secrets
- Review permissions regularly
- Document why each denial exists
❌ DON'T:
- Hardcode credentials in settings.json
- Use
allowmode for untrusted contexts - Grant
Bash(*)without restrictions - Include secrets in version control
- Mix personal and project settings
Permission Checklist
- All secrets use
${VAR_NAME}syntax - Dangerous patterns are denied
- File paths are explicit (not wildcards)
- Permission mode matches use case (ask/allow/deny)
- Hooks are not left in commented state
- MCP servers have proper OAuth configuration
- No
.envfile is readable - Sudo commands are denied
Reference: Claude Code settings.json documentation Version: 1.0.0