원클릭으로
config-audit
Audit Claude Code configuration for security: check permissions, hooks, MCP servers, and settings
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Audit Claude Code configuration for security: check permissions, hooks, MCP servers, and settings
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
Analyze, group, and commit uncommitted changes as atomic conventional commits. Use this skill whenever the user says 'commit', 'commit this', 'commit these changes', 'commit my changes', 'smart commit', wants their diff split into logical commits, or asks to create a commit in any way. Always use this skill for committing — never run git commit directly.
Create or update today's daily note at `~/knowledge/notes/journal/YYYY/MM-MMM/YYYY-MM-DD-DayName.md`, appending or replacing a session entry under `## Agent Summary`. Use when the user asks to update the daily note, log a session, append to the journal, or when invoked by the session-history hook after a session summary is written.
Edit config files inside the stow-based dotfiles repo and re-stow them with `make link`. Use this skill whenever the user asks to modify, add, or tweak any config under `$HOME` that lives in dotfiles — `.zshrc`, `.zshenv`, `.zprofile`, anything under `.config/`, `.local/`, `.claude/`, `.kiro/`, `.meshclaw/`, `.aws/`, `.unison/`, the `Makefile`, or any other tracked dotfile. Trigger even if the user phrases it as "change my config", "update my zshrc", "add this to claude settings", "tweak my tmux config", "modify my aliases", etc. — direct edits to `$HOME` are wrong because they bypass version control and get clobbered by the next `make link`.
Summarize unread emails via Outlook MCP, group by priority, draft replies
Audit IAM policies in CDK/CloudFormation/Terraform for least-privilege violations
Automatic security scanning on every code change with CWE/OWASP mapping
| name | config-audit |
| description | Audit Claude Code configuration for security: check permissions, hooks, MCP servers, and settings |
| effort | medium |
When asked to audit configuration, check security settings, or verify Claude Code setup:
Read settings — Check ~/.claude/settings.json for:
dangerouslySkipPermissions, bypassPermissions)Check MCP servers — Read ~/.claude/.mcp.json for:
autoApprove for sensitive toolsCheck hooks — Verify security hooks are active:
block-destructive-commands (PreToolUse:Bash)pre-write-security-scan (PreToolUse:Write|Edit).env, .ssh)Check CLAUDE.md — Verify expected behavioral guidelines are loaded.
Generate audit report
# Claude Code Security Audit
## Settings
- ✓ Destructive commands denied (rm -rf, sudo, git push)
- ✓ Sensitive file reads denied (.env, .ssh)
- ✗ bypassPermissions is enabled — DISABLE THIS
- ✓ Telemetry disabled
## Hooks (X/Y active)
- ✓ block-destructive-commands
- ✗ pre-write-security-scan — NOT INSTALLED
## MCP Servers
- ✓ servername: no autoApprove
- ⚠ custom-server: autoApprove includes write tools
## CLAUDE.md
- ✓ Behavioral guidelines loaded
## Recommendations
1. Disable bypassPermissions
2. Install pre-write-security-scan hook
3. Remove autoApprove from custom-server write tools