Skip to main content
command CX Codex official · Anthropic

brainstorm · anthropics — Claude Code command: what it does, how to install

brainstorm is a Claude Code command for exploring product ideas, problem spaces, and strategic questions. It frames the discussion, challenges assumptions, and helps narrow the options.

Apache-2.0 ★ 36,654 updated 2026-10-24

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add anthropics/knowledge-work-plugins
/plugin install product-management@knowledge-work-plugins
Open on GitHub

What brainstorm does

brainstorm turns a product question into a guided discussion. It asks one framing question, explores the problem or idea, then tests assumptions and considers alternative directions. The exchange is meant to develop the thinking, not produce a detached list of suggestions.

As the session progresses, the command moves between generating options and choosing among them. At the close, it captures the main ideas, a leading direction, its riskiest assumption, and a next step. Connected knowledge bases, analytics, project trackers, and team chat can provide context; otherwise, brainstorm works from the information the user shares.

What it helps with

  • explore a product problem from multiple angles
  • test assumptions behind a proposed feature
  • compare possible product strategies
  • identify a useful next step after an ideation session

Example prompts for the agent

  • /brainstorm Explore why users leave during onboarding and challenge my assumptions.
  • /brainstorm Test the marketplace idea and suggest other product directions.
  • /brainstorm Help frame our product strategy and identify its riskiest assumption.

How to install brainstorm

1Plugin marketplace

/plugin marketplace add anthropics/knowledge-work-plugins
/plugin install product-management@knowledge-work-plugins

2Manually: put the files in place

.claude/commands/brainstorm.md

Call it with /brainstorm. For all projects use ~/.claude/commands/

Questions about brainstorm

How do I install brainstorm in Claude Code?

Run /plugin marketplace add anthropics/knowledge-work-plugins, then run /plugin install product-management@knowledge-work-plugins. The command file is .claude/commands/brainstorm.md.

What does brainstorm do?

It guides a discussion about a product problem, idea, or strategic question. The command probes assumptions, develops alternatives, and closes with key ideas and a suggested next step.

Can brainstorm work without connected tools?

Yes. Without a knowledge base, analytics, project tracker, or team chat, it uses the context provided by the user.

Source text

brainstorm.md · Apache-2.0
---
description: Brainstorm a product idea, problem space, or strategic question with a sharp thinking partner
argument-hint: "<topic, problem, or idea to explore>"
---

# /brainstorm

> If you see unfamiliar placeholders or need to check which tools are connected, see [CONNECTORS.md](../CONNECTORS.md).

Brainstorm a product topic with a sharp, opinionated thinking partner. This is a conversation, not a deliverable — the goal is to push thinking further than the PM would get alone.

## Usage

```
/brainstorm $ARGUMENTS
```

## How It Works

```
┌────────────────────────────────────────────────────────────────┐
│                      BRAINSTORM                                │
├────────────────────────────────────────────────────────────────┤
│  STANDALONE (always works)                                     │
│  ✓ Explore problem spaces and opportunity areas                │
│  ✓ Generate and challenge product ideas                        │
│  ✓ Stress-test assumptions and strategies                      │
│  ✓ Apply PM frameworks (HMW, JTBD, First Principles, etc.)    │
│  ✓ Capture key ideas, next steps, and open questions           │
├────────────────────────────────────────────────────────────────┤
│  SUPERCHARGED (when you connect your tools)                    │
│  + Knowledge base: Pull prior research, specs, and decisions   │
│  + Analytics: Ground ideas in actual usage data                │
│  + Project tracker: Check what has been tried before           │
│  + Chat: Review recent team discussions for context            │
└────────────────────────────────────────────────────────────────┘
```

## Workflow

### 1. Understand the Starting Point

The PM might bring any of these — identify which one and adapt:

- **A problem**: "Our users drop off during onboarding" — start in problem exploration mode
- **A half-formed idea**: "What if we added a marketplace?" — start in assumption testing mode
- **A broad question**: "How should we think about AI in our product?" — start in strategy exploration mode
- **A constraint to work around**: "We need to grow without adding headcount" — start in solution ideation mode
- **A vague instinct**: "Something feels off about our pricing" — start in problem exploration mode

Ask one clarifying question to frame the session, then dive in. Do not front-load a list of questions. The conversation should feel like two PMs at a whiteboard, not an intake form.

### 2. Pull Context (if available)

If **~~knowledge base** is connected:
- Search for prior research, specs, or decision documents related to the topic
- Surface relevant user research findings or customer feedback
- Find previous brainstorming notes or exploration documents

If **~~product analytics** is connected:
- Pull relevant usage data, adoption metrics, or behavioral patterns
- Ground the brainstorm in real numbers rather than assumptions

If **~~project tracker** is connected:
- Check if similar ideas have been explored, attempted, or shelved before
- Look for related tickets, epics, or strategic themes

If **~~chat** is connected:
- Search for recent team discussions on the topic
- Surface relevant customer conversations or feedback threads

If these tools are not connected, work entirely from what the PM provides. Do not ask them to connect tools.

### 3. Run the Session

See the **product-brainstorming** skill for detailed guidance on brainstorming modes, frameworks, and session structure.

**Key behaviors:**
- Be a sparring partner, not a scribe. React to ideas. Push back. Build on them. Suggest alternatives.
- Match the PM's energy. If they are excited about a direction, explore it before challenging it.
- Use frameworks when they help, not as a checklist. If "How Might We" unlocks new thinking, use it. If the conversation is already flowing, do not interrupt with a framework.
- Push past the first idea. If the PM anchors on a solution early, acknowledge it, then ask for 3 more.
- Name what you see. If the PM is solutioning before defining the problem, say so. If they are stuck in feature parity thinking, call it out.
- Shift between divergent and convergent thinking. Open up when exploring. Narrow down when the PM has enough options on the table.
- Keep the conversation moving. Do not let it stall on one idea. If a thread is exhausted, prompt a new angle.

**Session rhythm:**
1. **Frame** — What are we exploring? What do we already know? What would a good outcome look like?
2. **Diverge** — Generate ideas. Follow tangents. No judgment yet.
3. **Provoke** — Challenge assumptions. Bring in unexpected perspectives. Play devil's advocate.
4. **Converge** — What are the strongest 2-3 ideas? What makes them interesting?
5. **Capture** — Document what emerged and what to do next.

### 4. Close the Session

When the conversation reaches a natural stopping point, offer a concise summary:

- **Key ideas** that emerged (2-5 ideas, each in 1-2 sentences)
- **Strongest direction** and why you think so — take a position
- **Riskiest assumption** for the strongest direction
- **Suggested next step**: the single most useful thing to do next (research, prototype, talk to users, write a one-pager, run an experiment)
- **Parked ideas**: interesting ideas that are worth revisiting but not right now

Do not generate the summary unprompted mid-conversation. Only summarize when the PM signals they are ready to wrap up, or when the conversation has naturally run its course.

### 5. Follow Up

After the session, offer:
- "Want me to turn the top idea into a one-pager?" → `/one-pager` or `/write-spec`
- "Want me to map this into an opportunity solution tree?"
- "Want me to draft a research plan to test the riskiest assumption?" → `/synthesize-research`
- "Want me to check how competitors approach this?" → `/competitive-brief`

## Tips

1. **This is a conversation, not a report.** Do not generate a 20-item idea list and hand it over. Engage with each idea. React. Build. Challenge.
2. **One good question beats five mediocre suggestions.** The right provocative question unlocks more than a list of options.
3. **Take positions.** "I think approach B is stronger because..." is more useful than presenting all options neutrally.
4. **Name the traps.** If you see the PM falling into feature parity thinking, solutioning before framing, or anchoring on constraints — say so directly.
5. **Know when to stop.** A brainstorm that goes too long produces fatigue, not ideas. If the PM has 2-3 strong directions and a clear next step, the session is done.

© Anthropic · Apache-2.0 · anthropics/knowledge-work-plugins

Similar slash commands

All slash commands for Notes, tasks and projects →

brainstorm by other authors comes first.

brainstorm SuperClaude-Org brainstorm is a command for Claude Code, Codex, GitHub Copilot, and Gemini CLI that clarifies vague project ideas and turns them into requirements before planning. CX Codex +1 ★ 23,950repo stars brainstorm Paweł Huryn brainstorm — a Claude Code command that produces product ideas or validation experiments from PM, designer and engineer viewpoints, for either an existing product or a new concept. CX Codex +1 ★ 28,211repo stars brainstorm-ideas-new phuryn brainstorm-ideas-new is a Claude Code skill that generates feature ideas from product manager, designer, and engineering perspectives for early product discovery and startup hypothesis validation. CX Codex +1 3,600installs bmad-brainstorm aj-geddes bmad-brainstorm runs structured ideation sessions using SCAMPER, SWOT, and 5 Whys. Delivers an organized markdown report ready for project planning. CX Codex +1 130installs brainstorm-ideas borghei brainstorm-ideas — a skill for structured product ideation: generates ideas from product, design, and engineering angles, maps each to a market opportunity, and identifies the top five. CX Codex +1 121installs game-brainstorm AlterLab-IEU game-brainstorm is a Claude Code skill that turns a theme or genre into three game concepts and a document for the selected idea. It starts from requests for a game idea, pitch, or game inspired by another title. CX Codex +1 ★ 57repo stars brainstorm-ideas-existing phuryn brainstorm-ideas-existing — a skill to brainstorm product ideas for an existing product from PM, Designer, and Engineer perspectives. Use it to generate features, solve identified opportunities, or ideate with your product trio. CX Codex +1 3,300installs bmad-brainstorming bmad-code-org The bmad-brainstorming skill runs a brainstorming session: the agent asks one question at a time, rotates creative techniques and logs ideas to a memlog. Useful when you want more options than you'd get alone. CX Codex +1 17,900installs

From the same set263

The whole set →
account-plan Builds or refreshes a strategic account plan by pulling data from your CRM, emails, transcripts, and notes. CX Codex +1 ★ 36,654repo stars account-context account-context is a Claude Code skill that gathers account details from connected sources into a brief. It runs when asked about a customer’s status or history. CX Codex +1 ★ 36,654repo stars start start is a Claude Code skill that initializes the productivity system and opens its dashboard. It is used during plugin setup or to bootstrap memory from tasks. CX Codex +1 ★ 36,654repo stars competitive-brief The competitive-brief Claude Code skill builds a brief on competitors or a product feature area. It supports product strategy, feature prioritization, sales preparation, and investor materials. CX Codex +1 ★ 36,654repo stars sign This Claude Code command places a signature or initials image in a PDF. It can use a detected signature field or a location specified for the page. CX Codex +1 ★ 36,654repo stars open open is a Claude Code command that displays PDFs in an interactive viewer. It runs when a document needs visual inspection rather than text extraction alone. CX Codex +1 ★ 36,654repo stars fill-form The fill-form Claude Code command fills PDF form fields in a live viewer. The agent checks unclear fields against page labels and shows the completed pages for review. CX Codex +1 ★ 36,654repo stars annotate This Claude Code command annotates PDFs with an agent. It proposes changes section by section, then shows each batch for review. CX Codex +1 ★ 36,654repo stars

More in the catalog

Project sponsors

Companies that keep this analytics open