Kipt

Brainstorming skill: what it does and how to install it

The brainstorming skill from Superpowers makes your coding agent ask what you want and why before it builds. What it does, its files, and how to install it.

By the Kipt team · Updated 24 October 2026

Source: obra/superpowers · Maintainer: community: Jesse Vincent (Prime Radiant) · License: MIT · Checked: 24 October 2026

What is the brainstorming skill?

The brainstorming skill makes a coding agent find out what you actually want, and why, before it writes any code. It asks one question at a time, plays its understanding back to you, lays out two or three ways to build the thing, and only moves on once you approve a written design.

It is the first step of Superpowers, the open source skill library by Jesse Vincent and Prime Radiant. The skill's description reads: "You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation)."

That wording is deliberately broad. The agent decides whether to load a skill from its description, so this one fires on almost any request to build or change something. Its skills.sh page listed about 403,900 installs (snapshot, 24 October 2026).

What does it do?

The SKILL.md opens with a blunt observation: the agent builds what it believes you want, "and that belief is usually thinner than it feels." Everything after that is a process for closing the gap.

First the agent sorts the request. If everything that matters is already in it ("make the icon cornflower blue"), it just does the work. If something non-trivial is missing, it starts a conversation. Before implementing, it must do four things in order:

  1. Lay out the approaches in a message of their own and have you pick one.
  2. Write the design doc, short for a small change or full for a project.
  3. Get your approval of that doc.
  4. Invoke the superpowers:writing-plans skill.

Along the way, the skill sets some firm habits:

  • Look before you ask. In an existing project, the agent reads the code and files you pointed at until it can explain how things work today. It should never ask what the project already answers.
  • One question per message. Each message ends with a single question, with no list of example answers attached. Open questions come first, because "Options steer."
  • Play it back. As it understands each part, it describes that part back in about 200 to 300 words and asks what is wrong or missing.
  • Defaults said out loud. Any choice the agent makes for you goes into the design as its call, never as something you agreed to.
  • A hard gate. Nothing gets built until you approve the design doc. Building includes product code, scaffolding, installing dependencies and invoking an implementation skill. Reading and exploring are fine.

Two exceptions to the gate: a throwaway spike you said yes to, and the case where you opt out of the questions.

For a project, the agent then runs a builder check: it sends a fresh subagent the design doc with the prompt in builder-check-prompt.md, asking what a builder would still need to know. It sorts the answers into what is already answered, minor calls it can make itself, and questions only you can answer, then brings you that last pile in one message.

What's in the folder:

  • SKILL.md: the process, the design doc outline, the hard gate and a red flags table.
  • builder-check-prompt.md: the prompt for the subagent that reviews the design.
  • visual-companion.md: how to run the optional browser companion for mockups and diagrams.
  • scripts/: start-server.sh, stop-server.sh, a Node.js server.cjs, helper.js and frame-template.html for that companion.

How do you install the brainstorming skill?

As part of Superpowers in Claude Code. The Superpowers README installs it from Anthropic's official marketplace:

/plugin install superpowers@claude-plugins-official

Plugin skills are namespaced as /plugin-name:skill-name per the Claude Code skills docs, so it shows up as /superpowers:brainstorming. The README also documents a second route through obra/superpowers-marketplace, plus installs for Codex, Cursor, Gemini CLI and other agents.

Just this skill. The skills CLI can copy one folder, exactly as shown on the skills.sh page:

npx skills add https://github.com/obra/superpowers --skill brainstorming

Add -g to install for all projects and -a claude-code to target Claude Code. The folder lands in .claude/skills/brainstorming/ in your project, or ~/.claude/skills/brainstorming/ with -g. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.

In the Claude apps. Because it is MIT licensed, you can zip the folder and upload it in Customize > Skills, as described in Anthropic's help article on skills. The visual companion's launch steps are written for coding agents such as Claude Code, Codex and Gemini CLI, so expect the chat version to stay in text.

Standalone, the skill still ends by invoking superpowers:writing-plans. Install the writing plans skill too if you want that hand-off to work.

How do you use it?

You usually don't call it. With Superpowers installed, the README's own check is to start a fresh session and ask "Let's make a react todo list"; a working install triggers brainstorming before any code is written. You can also name it, which tells the agent you want the full conversation.

Example prompts (suggestions, not quotes from the repo):

  • /superpowers:brainstorming I want a CLI that syncs my dotfiles across two laptops.
  • Add team invites to our app. Here's the spec in docs/invites.md, read it first.
  • Let's brainstorm a booking page for my yoga studio before we pick a stack.
  • Help me plan a 20 minute conference talk on feature flags.
  • Add CSV export to the reports page. (Without asking for brainstorming, expect one question plus an offer to just dive in.)

When you hand it a spec, the first message is meant to be about 200 words: your request echoed back, what it found in the code, anything that would change the plan, and one question.

Requirements: none for the conversation itself. The visual companion needs Node.js to run its local server. Its guide says mockups persist in your project's .superpowers/brainstorm/ folder and to add .superpowers/ to .gitignore.

Tips and limits

From the files, not from testing:

  • It is meant to slow the start down. On a vague request you will get questions before code. If you only want a quick change, take the "just dive in" offer or say so up front; Superpowers says your own instructions take precedence over its skills.
  • Approval means the doc. The red flags table is explicit: "sounds good" to an idea is not approval of a design doc the agent hasn't written yet.
  • Spikes stay throwaway. Keeping what a spike produced is treated as a new request that goes back through the gate.
  • What runs. Only the visual companion's scripts, and only after you accept the offer. They start a local web server that serves mockups over a URL with a session key, and it shuts down after 4 hours idle by default.
  • Audit note. On 24 October 2026, skills.sh showed Gen Agent Trust Hub and Socket as Pass and Snyk as Warn (medium risk, audited 22 October 2026), because the companion fetches a branding logo from primeradiant.com at runtime. That is the telemetry the README discloses; setting SUPERPOWERS_DISABLE_TELEMETRY turns it off.
  • License. MIT, so you can copy and edit the skill if you keep the copyright notice.

How is it different from grill-me?

Matt Pocock's grill me skill also interviews you before you build, but it is a short, on-demand interrogation of a plan until every decision is settled. Superpowers brainstorming runs by default on any creative request, adds a choice of approaches, writes a committed design doc, has a subagent review it, and then hands off to planning. If you want a quick stress test of an idea, grill-me is lighter. If you want design as a fixed first step of every feature, use brainstorming.

Save it in your Kipt library

In Kipt, open Skills and paste the skill's GitHub folder link into the import box. Kipt reads the SKILL.md and its files, including the builder check prompt and the companion scripts; for a repo with many skills it shows a list to pick from, so you can take writing-plans at the same time. You can also drop the folder or a .zip. Your skills live in your own private GitHub repo and install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Jesse Vincent, Prime Radiant or Superpowers.

Similar skills

Questions

What is the brainstorming skill?

brainstorming is an Agent Skill in Jesse Vincent's Superpowers library (obra/superpowers, MIT License). It makes a coding agent find out what you want and why before it builds anything: it asks one question at a time, lays out two or three approaches, writes a design doc, and waits for your approval. Then it hands off to the writing-plans skill.

How do I install the brainstorming skill in Claude Code?

Install the whole Superpowers plugin with /plugin install superpowers@claude-plugins-official, or copy only this skill with npx skills add https://github.com/obra/superpowers --skill brainstorming. The plugin version runs as /superpowers:brainstorming.

Can I skip the questions and just have the agent build?

Yes. When you did not ask for brainstorming, the skill tells the agent to add a one-line offer to just get going. If you take it, the skill is done: the agent states any big choice such as platform or language in one line and then does what you asked.

Where does the brainstorming skill save the design doc?

In a software repo, the SKILL.md says to commit it at docs/superpowers/specs/YYYY-MM-DD-<topic>-design.md. Outside a repo, the agent asks you where it should go, and your own preferences override both.

Does the brainstorming skill send any data?

Only its optional visual companion does. The Superpowers README says the companion loads the Prime Radiant logo from their website, which includes the Superpowers version and nothing about your project. Set SUPERPOWERS_DISABLE_TELEMETRY to any true value to turn it off.

Is it only for software?

No. Its description names non-software work too: a talk, a business, a renovation. The design doc location and the builder check are written for repos, but the questions and approach steps work for any project.

More in Skill directory

All guides

Keep your prompts and skills in one place, in your own GitHub. Install every skill with one command.

Get started

Kipt is an independent product, not affiliated with or endorsed by Anthropic, OpenAI, Cursor or GitHub. Claude, ChatGPT, Codex, Cursor and GitHub are trademarks of their owners.