원클릭으로
update-config
Configure the agent via settings.json — permissions, hooks, env vars, and tool rules
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Configure the agent via settings.json — permissions, hooks, env vars, and tool rules
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 | update-config |
| description | Configure the agent via settings.json — permissions, hooks, env vars, and tool rules |
| when_to_use | Use when the user wants to configure automated behaviours ("from now on when X", "each time X"), grant or revoke tool permissions ("allow bash", "deny write to /etc"), set environment variables, or modify .claude/settings.json and ~/.claude/settings.json. For simple preferences (model choice, verbosity) use the Config tool instead. |
| argument-hint | what to configure, e.g. 'allow npm commands' or 'run go test before committing' |
| context | inline |
| allowed-tools | ["read_file"] |
Help the user configure the agent by editing the appropriate settings file.
$ARGUMENTS
| Scope | Path | Who sees it |
|---|---|---|
| User (global) | ~/.claude/settings.json | You, all projects |
| Project | .claude/settings.json | Everyone in the repo (commit this) |
| Local | .claude/settings.local.json | You only, this project (gitignore this) |
Priority: local → project → user (local wins on conflict).
{
// Tool permission rules — evaluated top-to-bottom, first match wins.
// Each rule applies to one tool. Use "*" to match all tools.
"permissions": {
"allow": [
"bash(npm *)", // allow npm commands
"bash(go test *)", // allow go test
"write_file(src/**)" // allow writes under src/
],
"deny": [
"bash(rm -rf *)", // deny dangerous rm
"write_file(/etc/**)" // deny writes to /etc
]
},
// Environment variables injected into every tool execution.
"env": {
"GOFLAGS": "-mod=vendor",
"NODE_ENV": "development"
},
// Hooks — shell commands the harness runs on specific events.
// Claude itself does NOT run these; the harness does.
"hooks": {
// Fires before every tool call. Exit non-zero to block the call.
"PreToolUse": [
{
"matcher": "bash", // tool name pattern (glob)
"hooks": [
{ "command": "echo 'about to run bash'" }
]
}
],
// Fires after every tool call.
"PostToolUse": [
{
"matcher": "write_file",
"hooks": [
{ "command": "gofmt -w $CLAUDE_FILE_PATH 2>/dev/null || true" }
]
}
],
// Fires when the agent finishes a turn.
"Stop": [
{
"hooks": [
{ "command": "notify-send 'Agent done'" }
]
}
]
}
}
Allow a tool or command family:
{ "permissions": { "allow": ["bash(npm *)"] } }
Block a dangerous pattern:
{ "permissions": { "deny": ["bash(rm -rf *)"] } }
Auto-format Go files after every write:
{
"hooks": {
"PostToolUse": [
{ "matcher": "write_file", "hooks": [{ "command": "gofmt -w $CLAUDE_FILE_PATH || true" }] }
]
}
}
Run tests before every commit:
{
"hooks": {
"PreToolUse": [
{ "matcher": "bash(git commit*)", "hooks": [{ "command": "go test ./..." }] }
]
}
}
Set a project-wide environment variable:
{ "env": { "DATABASE_URL": "postgres://localhost/myapp_dev" } }
Research and plan a large-scale mechanical change, then execute it in parallel across independent work units
Review staged changes and create a conventional git commit
Diagnose agent session issues — show recent logs, tool errors, and configuration problems
Review auto-memory entries and promote useful ones to CLAUDE.md or project memory files
Review changed code for reuse, quality, and efficiency, then fix any issues found
Capture the current session's repeatable process into a reusable SKILL.md file