Claude Code Command Selection Guide

This skill helps you choose the most appropriate command, agent, or skill for different scenarios.

Quick Decision Flowchart

graph TD
    A[User Request] --> B{Request Type?}
    B -->|New Feature| C[/plan]
    B -->|Bug Fix| D[/tdd or build-error-resolver]
    B -->|Code Review| E[/code-review or code-reviewer agent]
    B -->|Testing| F[/e2e or tdd-guide agent]
    B -->|Context Too Long| G[/compact]
    B -->|Documentation| H[/docs or docs-lookup agent]
    B -->|Looping Task| I[/loop]
    B -->|Security Review| J[security-reviewer agent]

    C --> K[planner agent]
    D --> L{Build Failed?}
    L -->|Yes| M[build-error-resolver]
    L -->|No| N[tdd-guide]
    E --> O[code-reviewer]
    F --> P[e2e-runner]

1. Built-in Slash Commands

Session Management Commands

Development Workflow Commands

Documentation & Query Commands

2. Agents Selection

Development Workflow Agents

Problem Solving Agents

Research & Exploration Agents

3. Skills Selection

Workflow Skills

Architecture & Pattern Skills

Testing Skills

Research Skills

4. Scenario Decision Matrix

By Task Phase

By Problem Type

By Development Type

5. Parallel Execution Strategy

Parallelizable Scenarios

Recommended: Launch multiple independent tasks simultaneously

Scenario: Preparing PR after code completion

• Agent 1: code-reviewer (code quality)
• Agent 2: security-reviewer (security review)
• Agent 3: e2e-runner (E2E tests)

Scenario: Large refactor analysis

• Agent 1: architect (architecture analysis)
• Agent 2: Explore (code exploration)
• Agent 3: refactor-cleaner (dead code detection)

Sequential Execution Required

Cannot parallelize: Dependencies exist

Scenario: Fixing build error

• Sequence: build-error-resolver -> test verification -> code-reviewer

Scenario: New feature development

• Sequence: planner -> tdd-guide (write tests) -> implementation -> code-reviewer

6. Auto-Trigger Rules

Invoke Without User Request

7. Context Management Timing

Best Practices:

• Compact after research, before implementation (preserve plan)
• Compact after milestone completion (clear intermediate state)
• Don't compact mid-implementation (lose variables/paths)

8. Command Cheat Sheet

Development Workflow:
/plan        -> Enter planning mode (complex tasks)
/tdd         -> TDD workflow
/e2e         -> E2E testing
/code-review -> Code review
/build-fix   -> Fix build

Session Management:
/compact     -> Compact context
/clear       -> Clear session
/loop        -> Looping task
/fast        -> Fast mode

Documentation & Memory:
/docs        -> Update docs
/remember    -> Save memory
/tasks       -> View tasks

Help:
/help        -> View all commands

9. Usage Examples

Example 1: New Feature Development

User: Add user authentication feature

Workflow:

1. /plan -> planner agent creates plan
2. tdd-guide -> write tests
3. Implementation -> edit code
4. code-reviewer -> code review
5. security-reviewer -> security review (auth sensitive)
6. e2e-runner -> E2E tests
7. /compact -> compact after milestone completion

Example 2: Build Failure

User: npm run build failed

Workflow:

1. build-error-resolver -> analyze error, minimal fix
2. Verify build success
3. code-reviewer -> check fix quality

Example 3: Code Refactoring

User: Refactor authentication module

Workflow:

1. architect -> architecture analysis
2. planner -> implementation plan
3. refactor-cleaner -> dead code detection
4. tdd-guide -> ensure test coverage
5. Implementation -> refactor code
6. verification-loop -> comprehensive verification

Core Principles:

1. Plan first, implement later - Use /plan for complex tasks
2. Test first - Use tdd-guide for new features
3. Review immediately after coding - Use code-reviewer when code complete
4. Fix build immediately when failed - Use build-error-resolver
5. Review sensitive code - Use security-reviewer for auth/API
6. Verify comprehensively before PR - Use verification-loop