Command Selection Guide¶

User Reference: This guide helps you choose the right slash command for your workflow.

Quick Command Finder¶

Decision Tree:

What do you need to do?
├─ DEVELOPMENT WORKFLOWS
│  ├─ Quick fix (< 5 min)? → /fix
│  ├─ Build new module? → /modular-build
│  ├─ Full feature development? → /ultrathink
│  └─ Autonomous external loop? → /auto
│     └─ Note: /auto runs in subprocess, /ultrathink in current session
│
├─ DECISION MAKING
│  ├─ Need discussion/debate? → /debate
│  │  └─ Interactive facilitated discussion converging to consensus
│  └─ Need expert voting? → /expert-panel
│     └─ Independent reviews with weighted voting
│
├─ FAULT TOLERANCE
│  ├─ Need reliability/fallbacks? → /cascade
│  ├─ Need critical correctness? → /n-version
│  └─ Need decision consensus? → /debate
│
└─ INVESTIGATION & IMPROVEMENT
   ├─ Understand codebase? → /investigate
   ├─ Session reflection? → /reflect
   └─ Philosophy compliance? → /analyze

Quick Reference Table¶

Task Type	Command	When to Use
Quick fixes	`/fix [pattern] [scope]`	Error blocking you, need rapid resolution
Module building	`/modular-build`	Creating new self-contained module (brick)
Feature development	`/ultrathink`	Multi-step feature with workflow orchestration
Autonomous work	`/auto`	Long-running task in external subprocess
Debate decision	`/debate <question>`	Need facilitated discussion to consensus
Expert consensus	`/expert-panel <topic>`	Need independent expert votes
Fallback resilience	`/cascade <task>`	Need graceful degradation (API calls, etc.)
Critical correctness	`/n-version <task>`	Security code, core algorithms (3-4x cost)
Code understanding	`/investigate <path>`	Deep dive into codebase structure
Session analysis	`/reflect`	Review session and create improvement issues
Philosophy check	`/analyze <path>`	Validate code against amplihack principles

Key Distinctions¶

`/ultrathink` vs `/auto`¶

/ultrathink: Runs in current Claude Code session (internal orchestration)
/auto: Spawns external subprocess with autonomous loop
Use /ultrathink for: Interactive development, want to see progress
Use /auto for: Hands-off background work, long-running tasks

`/debate` vs `/expert-panel`¶

/debate: Interactive discussion with back-and-forth, facilitated convergence
/expert-panel: Independent reviews without cross-talk, then voting
Use /debate for: Exploring trade-offs, need discussion
Use /expert-panel for: Clear decision with expert consensus

`/ultrathink` vs `/modular-build`¶

/ultrathink: General-purpose development workflow
/modular-build: Specialized for module creation (progressive pipeline)
Use /ultrathink for: Features, bug fixes, general development
Use /modular-build for: Specifically creating new bricks (self-contained modules)

Command Integration Examples¶

Quick Workflow: Fix → Build → Review¶

/fix import                    # Fix import errors
/modular-build auth-module     # Build new module
/analyze .                     # Check philosophy compliance

Complex Decision Workflow: Debate → N-Version → Reflect¶

/debate "Should we use REST or GraphQL?"
/n-version "Implement chosen API approach"
/reflect                       # Analyze decisions made

Investigation Workflow: Investigate → Ultrathink¶

/investigate ./src/core        # Understand existing code
/ultrathink "Add new feature to core"  # Build on understanding

All Available Commands¶

Development Commands¶

/fix - Quick error resolution with pattern detection
/modular-build - Create self-contained modules (bricks)
/ultrathink - Full development workflow
/auto - Autonomous external subprocess

Decision Making Commands¶

/debate - Facilitated multi-agent discussion
/expert-panel - Independent expert review with voting

Fault Tolerance Commands¶

/cascade - Graceful degradation with fallbacks
/n-version - N-version programming for critical code

Investigation Commands¶

/investigate - Deep codebase exploration
/analyze - Philosophy compliance check
/reflect - Session analysis and improvement

Document-Driven Development (DDD)¶

/ddd:0-help - DDD workflow guide
/ddd:prime - Load DDD context
/ddd:1-plan - Planning phase
/ddd:2-docs - Documentation retcon
/ddd:3-code-plan - Code planning
/ddd:4-code - Implementation
/ddd:5-finish - Cleanup and finalize
/ddd:status - Check DDD progress

Utility Commands¶

/customize - Manage user preferences
/skill-builder - Create new Claude Code skills
/knowledge-builder - Build knowledge bases
/socratic - Generate Socratic questions
/transcripts - Manage conversation transcripts
/lock / /unlock - Continuous work mode
/install / /uninstall - System setup

When in Doubt¶

Start with /ultrathink - it will orchestrate the appropriate workflow for most tasks.

For quick questions or simple tasks, just ask directly without commands.

See Also:

.claude/context/AGENT_SELECTION_GUIDE.md - Which agent to use (AI context)
CLAUDE.md - Complete project documentation
.claude/commands/ - Individual command documentation