This page documents the commands/ directory: the three command definition files that expose superpowers skills as named slash commands in Claude Code. It covers the file format, the disable-model-invocation flag, and how each command maps to its backing skill.
For documentation on the skills these commands invoke, see the Skills Reference (7). For the hooks system that injects context at session start, see 10.3. For directory layout context, see 10.1.
Claude Code supports custom slash commands defined as Markdown files placed in a commands/ directory. The superpowers plugin uses this mechanism to surface three high-frequency workflow entry points as first-class commands, so users can invoke them directly (e.g., /brainstorm) rather than describing what they want in natural language.
Each command file is intentionally minimal: its only job is to load and delegate to a superpowers skill. No workflow logic lives in the command files themselves.
| File | Slash Command | Skill Invoked | Description Field |
|---|---|---|---|
commands/brainstorm.md | /brainstorm | superpowers:brainstorming | "You MUST use this before any creative work..." |
commands/write-plan.md | /write-plan | superpowers:writing-plans | "Create detailed implementation plan with bite-sized tasks" |
commands/execute-plan.md | /execute-plan | superpowers:executing-plans | "Execute plan in batches with review checkpoints" |
Sources: commands/brainstorm.md1-7 commands/write-plan.md1-7 commands/execute-plan.md1-7
Each command file is a Markdown file with a YAML frontmatter block followed by a plain-text body.
General structure:
---
description: <human-readable description shown in the command picker>
disable-model-invocation: true
---
Invoke the superpowers:<skill-name> skill and follow it exactly as presented to you
| Field | Type | Purpose |
|---|---|---|
description | string | Shown in Claude Code's command picker UI. Also used for discoverability matching. |
disable-model-invocation | boolean | When true, prevents Claude from generating a freeform response before executing the command body. Forces deterministic behavior. |
The disable-model-invocation: true flag is present in all three command files. Without it, the model might respond to the user's invocation with its own interpretation before running the skill, which could contradict or bypass the skill's structured workflow.
Sources: commands/brainstorm.md1-4 commands/write-plan.md1-4 commands/execute-plan.md1-4
commands/brainstorm.md/brainstormsuperpowers:brainstormingcommands/write-plan.md/write-plansuperpowers:writing-planscommands/execute-plan.md/execute-plansuperpowers:executing-plansThe body of every command file is a single instruction:
Invoke the superpowers:<skill-name> skill and follow it exactly as presented to you
This uses the superpowers: namespace prefix (documented in 3.3) to explicitly target the superpowers skill library rather than any project-local or personal override. The phrase "follow it exactly" is a guard against the model selectively interpreting or abbreviating the skill's steps.
Delegation flow diagram:
Sources: commands/brainstorm.md6-7 commands/write-plan.md6-7 commands/execute-plan.md6-7
Commands are defined for the first three stages of the mandatory development workflow. They are the user-facing entry points into that pipeline.
The full workflow pipeline (including stages that do not have dedicated commands, such as subagent-driven development and finishing a branch) is documented in 6.1.
Sources: commands/brainstorm.md1-7 commands/write-plan.md1-7 commands/execute-plan.md1-7
The commands expose a convenience layer only. The following are not defined as commands:
subagent-driven-development — this is invoked by the agent internally, not by the user directly.finishing-a-development-branch — invoked by the agent at the end of the workflow.systematic-debugging, test-driven-development, and other technique skills — these are invoked by the agent when conditions match, not by user slash commands.For skills that are invoked automatically rather than by explicit user command, see 3.2 (Mandatory Skill Check Protocol) and 3.3 (Finding and Invoking Skills).
Refresh this wiki
This wiki was recently refreshed. Please wait 3 days to refresh again.