Command Patterns Skill

Expert patterns for generating production-quality Claude Code commands.

Core Requirements

1. Multiple Action Modes - Never single-purpose commands

   • analyze - Deep analysis
   • optimize - Performance improvements
   • fix - Automated remediation
   • monitor - Continuous watching

2. Real, Working Code - Every example must be executable

3. Intelligent Recommendations - Prioritized, actionable insights

4. Interactive Intelligence - Guide users to optimal solutions

5. Visual Outputs - Include diagrams/charts when helpful

Advanced Patterns

Progressive Disclosure

# Simple for beginners
claude /your-command

# Power user with full control
claude /your-command analyze --depth=comprehensive --output=json

Safe Automation with Rollbacks

apply_change() {
  # Snapshot current state
  kubectl get deployment $1 -o yaml > "/tmp/rollback-$$.yaml"

  # Apply change
  kubectl patch deployment $1 --patch "$2"

  # Monitor and rollback if needed
  if ! monitor_health $1 60; then
    kubectl apply -f "/tmp/rollback-$$.yaml"
  fi
}

Quality Checklist

Intelligence

• [ ] Recognizes patterns from past executions
• [ ] Provides insights humans would miss
• [ ] Explains reasoning behind recommendations

User Experience

• [ ] Works with zero configuration
• [ ] Provides meaningful progress indicators
• [ ] Offers escape hatches at every step

Safety

• [ ] Every change has a rollback
• [ ] Dry-run mode available
• [ ] Confirms destructive operations

Integration

• [ ] Outputs in multiple formats
• [ ] Works in CI/CD pipelines
• [ ] Provides hooks for customization

Frontmatter Template

---
description: Brief description  # REQUIRED
model: haiku                    # OPTIONAL - use full ID
allowed-tools: Bash, Read       # OPTIONAL
argument-hint: "[action]"       # OPTIONAL
---

The Ultimate Test

"Would a senior engineer with 10 years experience be impressed?"

If not emphatic YES, keep improving.