Skip to content

Claude Code Configuration

This infrastructure includes comprehensive Claude Code enhancements for AI-assisted operations.

Overview

Claude Code is configured with custom skills, agents, commands, and safety hooks specific to this infrastructure. These enhancements enable:

  • Context-aware assistance - Claude understands this specific infrastructure
  • Safety guardrails - Dangerous commands are blocked before execution
  • Specialized agents - Domain-specific experts for different tasks
  • Quick commands - Slash commands for common operations

Components Summary

Type Count Location
Skills 12 .claude/skills/
Custom Agents 4 .claude/agents/
Quick Commands 4 .claude/commands/infra/
Safety Hooks 3 .claude/hooks/

Skills

Skills are model-invoked capabilities that activate automatically based on context.

Infrastructure-Specific Skills

Skill Activates When Purpose
monitoring-operations Discussing Prometheus, Grafana, Loki PromQL queries, LogQL, dashboards
notification-management Working with ntfy, Apprise Push notifications, SMTP gateway
security-auditing Security reviews, SSL checks Certificate verification, secret detection
git-operations Gitea, GitHub sync Repository management, BFG filtering
redis-management Redis operations cache-redis, session-redis management
media-services Jellyfin, Metube Media streaming and downloads

Core Skills (Pre-existing)

Skill Purpose
backup-restore Backup procedures and disaster recovery
database-operations PostgreSQL and Supabase database ops
docker-stack-manager Docker Compose stack management
infrastructure-diagnostics Troubleshooting and health checks
n8n-workflow-builder n8n automation workflows
service-deployment New service deployment with Traefik

Custom Agents

Specialized agents with infrastructure-specific knowledge.

infrastructure-specialist

Model: Sonnet | Tools: Read, Grep, Glob, Bash, WebFetch

Context-aware expert on this specific infrastructure. Knows: - Service topology and dependencies - Network architecture (traefik_proxy, supabase_default) - Configuration file locations - Critical service paths

media-manager

Model: Haiku | Tools: Read, Bash, Glob

Fast operations for media services: - Jellyfin library management - Metube download troubleshooting - Stremthru and zurg configuration

security-reviewer

Model: Sonnet | Tools: Read, Grep, Glob (read-only)

Security auditing without modification capability: - Configuration review - Secret detection - SSL/TLS verification - Container security analysis

log-analyzer

Model: Haiku | Tools: Read, Grep, Glob, Bash

Fast log analysis and correlation: - Error pattern detection - Cross-service correlation - Incident timeline building

Quick Commands

Slash commands for common infrastructure operations.

/infra/quick-status     - Fast infrastructure health overview
/infra/service-health   - Detailed service diagnostics
/infra/emergency-stop   - Safe shutdown procedures
/infra/backup-now       - Trigger immediate backup

Using Commands

Commands are invoked with the slash prefix: 

/infra/quick-status

Some commands accept arguments: 

/infra/service-health grafana
/infra/emergency-stop non-critical
/infra/backup-now infrastructure

Safety Hooks

Hooks intercept tool execution to enforce safety policies.

PreToolUse Hooks

dangerous-command-blocker.py Blocks catastrophic commands before execution: - rm -rf / and similar patterns - docker system prune -a - DROP DATABASE, TRUNCATE TABLE - Writing to system directories (/etc, /usr)

production-service-guard.py Requires confirmation for critical service operations: - Protected services: supa-db, traefik, prometheus, grafana, alertmanager, loki, authentik, adguard, gitea, cache-redis, session-redis - Protected operations: stop, remove, kill, restart, down

PostToolUse Hook

audit-logger.py Logs all tool executions for audit trail: - Location: .claude/logs/audit.log - Includes: timestamp, session ID, tool name, input summary, result status

Configuration Files

File Purpose
.claude/settings.local.json Hook registration, permissions, MCP servers
.claude/skills/*/SKILL.md Individual skill definitions
.claude/agents/*.md Agent configurations
.claude/commands/infra/*.md Command definitions

Viewing Audit Logs

# View recent audit entries
tail -100 /srv/dockerdata/.claude/logs/audit.log

# Search for specific tool usage
grep "Bash" /srv/dockerdata/.claude/logs/audit.log | tail -20

# View today's activity
grep "$(date +%Y-%m-%d)" /srv/dockerdata/.claude/logs/audit.log

Extending the Configuration

Adding a New Skill

  1. Create directory: .claude/skills/my-skill/
  2. Create SKILL.md with YAML frontmatter: 
    ---
name: my-skill
description: Description of when this skill should activate
---

# Skill Name

## Purpose
[What this skill enables]

## Operations
[Commands and procedures]

Adding a New Agent

Create .claude/agents/my-agent.md: 

---
name: my-agent
description: What this agent specializes in
tools: Read, Grep, Glob
model: haiku
---

# Agent Name

## Your Responsibilities
[What this agent should focus on]

## Key Knowledge
[Infrastructure-specific context]

Adding a New Command

Create .claude/commands/category/my-command.md: 

---
description: What this command does
arguments:
  - name: arg1
    description: Optional argument
    required: false
---

# Command Name

[Instructions for Claude to follow when command is invoked]