Skip to content

Command Reference: Slash Commands Quick Reference ​

Superpowers provides three slash commands for quickly invoking core workflow skills.

Command Overview ​

CommandDescriptionCalls SkillUse Case
/brainstormCreative design requirement clarificationbrainstormingBefore creative work
/write-planWrite detailed implementation planwriting-plansBefore multi-step tasks with specs/requirements
/execute-planBatch execute plan tasksexecuting-plansWhen executing an existing plan

/brainstorm ​

Description: Clarify requirements and explore creative ideas before design

Trigger Conditions:

  • Create new feature
  • Build new component
  • Add new functionality
  • Modify existing behavior

Calls Skill: superpowers:brainstorming

Usage Example:

bash
/brainstorm "Add user profile editing feature"

Core Process:

  1. Understand project context
  2. Ask clarification questions one at a time
  3. Present design proposal (in 200-300 word segments)
  4. Save design document to docs/plans/
  5. Commit design to version control

Important Notes:

  • Must be executed before writing code
  • Ask only one question at a time
  • Design document must be committed

/write-plan ​

Description: Break down requirements into executable small tasks

Trigger Conditions:

  • Clear requirements or specifications exist
  • Tasks require multi-step implementation
  • Need detailed implementation plan

Calls Skill: superpowers:writing-plans

Usage Example:

bash
/write-plan "docs/plans/2024-02-01-user-profile-design.md"

Core Process:

  1. Read design document or requirement specifications
  2. Break down into 2-5 minute tasks
  3. Write complete code for each task
  4. Include testing steps
  5. Save implementation plan

Task Breakdown Principles:

  • Each task completable in 2-5 minutes
  • Task granularity small enough for "no-context" engineers to execute
  • Each task includes implementation and verification

Important Notes:

  • Tasks must be sufficiently small
  • Each task must have acceptance criteria
  • Plan document should be clear and readable

/execute-plan ​

Description: Batch execute tasks from a plan with review checkpoints

Trigger Conditions:

  • Implementation plan is already written
  • Need to batch execute tasks
  • Execute in separate session (recommended)

Calls Skill: superpowers:executing-plans

Usage Example:

bash
/execute-plan "docs/plans/2024-02-01-user-profile-implementation.md"

Core Process:

  1. Read implementation plan
  2. Execute tasks one by one
  3. Run tests after each task
  4. Set review checkpoints after each batch
  5. Verify after completing all tasks

Execution Strategy:

  • Execute tasks in order
  • Verify tests after each task
  • Conduct code review after each batch (typically 3-5 tasks)
  • Stop and debug on failure

Important Notes:

  • Suitable for execution in separate sessions
  • Difference from subagent-driven-development is in execution approach
  • Must follow TDD principles

Command Usage Recommendations ​

When to Use Commands ​

Use commands when:

  • You know exactly which workflow you need
  • Want to quickly start a standard process
  • Using Claude Code

Use skills directly when:

  • Unsure which skill is needed
  • AI proactively suggests using a skill
  • Calling through Skill tool

Commands vs Skills ​

AspectSlash CommandsSkill Calls
Invocation MethodUser actively inputs /commandAI auto-invokes or through Skill tool
Use CaseYou know which process is neededLet AI recommend appropriate process
Platform SupportClaude CodeClaude Code, OpenCode, Codex

Typical Workflows ​

Complete Development Process:

  1. /brainstorm - Design requirements
  2. /write-plan - Write plan
  3. /execute-plan - Execute implementation

Bug Fix Process:

  • Use systematic-debugging skill directly

Quick Prototype Process:

  • /brainstorm - Quick design
  • Implement directly (skip detailed planning)

Chapter Summary ​

This chapter introduced three slash commands provided by Superpowers:

  • /brainstorm: Creative design requirement clarification
  • /write-plan: Write detailed implementation plan
  • /execute-plan: Batch execute plan tasks

These commands provide convenient access to core workflows in Claude Code, suitable for scenarios where you know which process you need.

Next Lesson Preview ​

In the next lesson, we'll learn Agent Reference.

You'll learn:

  • Concepts and purposes of the agent system
  • Detailed explanation of code-reviewer agent
  • How to define and use custom agents

Appendix: Source Code Reference ​

Click to expand source code locations

Updated: 2026-02-01

CommandFile PathLine
/brainstorm command definitioncommands/brainstorm.md1-7
/write-plan command definitioncommands/write-plan.md1-7
/execute-plan command definitioncommands/execute-plan.md1-7

Key Features:

  • disable-model-invocation: true: Prevents direct model invocation, ensuring execution through the skill system
  • Command files only contain instruction calls; actual logic is in corresponding skills

Invocation Chain:

Slash command → Skill system (Skill tool) → Corresponding skill (SKILL.md) → Execute workflow