Before you write a plan, you need clarity on what you're actually trying to do. That's what brainstorming is for. Claude Code has specific skills designed to help with this phase - the most useful one is /brainstorming from the superpowers pack.

What good brainstorming produces

A clear statement of

• The goal - what you're actually trying to achieve (not what you think you need to build)
• Success criteria - how you'll know when it's done
• Constraints - what the work can't involve, what budget/timeline limits apply
• The shape - roughly what the output will look like
• Known unknowns - what you need to figur e out during the work

This becomes your brief.md - the anchor document the whole project references.

Before you start - install the superpowers pack

This module (+ the next 4 lessons) uses the superpowers skills pack - a community-maintained bundle that includes /brainstorming, /writing-plans, /executing-plans, and several others. If you installed it back in Module 09 Lesson 2, skip ahead. If not:

Install the superpowers pack from github.com/obra/superpowers. Confirm /brainstorming and /writing-plans are available as slash commands.

Restart Claude Code after install so the skills load. Takes 2 minutes.

Using the brainstorming skill (from the superpowers pack)

terminal
/brainstorming

I want to build a quarterly growth report for my top client. It should show what worked, what didn't, and propose a plan for next quarter. I've never built something like this before. Help me clarify scope.

The skill asks you 10-15 clarifying questions one at a time. Covers:

• Audience - who will read this?
• Decisions - what decisions will they make from it?
• Current state - what do they know already?
• Format - written report, presentation, dashboard?
• Constraints - budget, timeline, data access, confidentiality?
• Precedent - have you or they done similar before? What worked/didn't?

At the end, you have a written brief that captures the real shape of the work.

Without the skill - manual brainstorming

Same pattern, just prompt-driven

I want to [describe goal]. Help me brainstorm the scope. Ask me 10 questions one at a time that will force me to clarify what I'm actually trying to do. Don't propose solutions yet

• just help me think clearly about the problem.

Claude asks questions. You answer. At the end

Based on my answers, write a brief.md that captures the goal, success criteria, constraints, known unknowns, and rough output shape. Save to the project folder.

Anti-patterns

Jumping to solutions before understanding the problem

"Build me a dashboard" - instead of - "what decisions does this dashboard need to support?" Always answer the second question first. The first shape usually isn't the right shape.

Over-specifying before you know what's hard

The most interesting brainstorming surfaces the 2-3 unknowns that will shape the whole project. If you lock in specifications before discovering those, you'll be re-working later.

Brainstorming alone vs with Claude

You CAN brainstorm alone in your head. But Claude's job is to poke the assumptions you didn't notice. Brainstorming with Claude surfaces more than solo thinking.

Skipping brainstorm because you "know what you want"

If you actually know what you want with full clarity, a 5-minute brainstorm costs you nothing and confirms it. If it uncovers something you missed - you just saved hours.

What a good brief.md looks like

Q1 Growth Report for Client Acme

Goal

Help the client make 3 decisions

1. Should they continue investing in Meta Ads at current levels?
2. Is their new SEO investment starting to pay off?
3. What's the single biggest opportunity for Q2?

Success criteria

• 3 decisions above have clear data-backed recommendations
• Report is readable in under 15 minutes by the CEO
• No unnecessary metrics - every number drives a decision

Constraints

• Delivered by [date]
• Uses data from our Supabase warehouse only (no manual pulls)
• Encrypted before publishing (client-facing)
• Brand guide: use the client's colours + logo, not ours

Known unknowns

• Whether SEO organic traffic gains are from actual ranking or algo changes (need to check)
• How to handle the Meta performance drop in Feb (attribution issue or real decline?)
• Whether the client wants a written report or presentation deck

Rough output shape

HTML report, 6-8 sections, embedded charts, encrypted URL. Template from my existing weekly-report format.

Specific enough to guide work. General enough to accommodate discoveries.

Power-user tips

• Write the brief in your voice - don't let Claude over-polish it. Rough bullet points beat clean prose.
• List known unknowns explicitly - knowing what you don't know is half the work
• Time-box brainstorming to 20 minutes - longer usually = procrastination
• Save the brief to the project folder as brief.md - Claude references it throughout the project
• Revisit the brief weekly on long projects - scope drifts, the brief is the anchor. If you catch yourself doing this regularly, turn it into a /brief-review skill using the skill-creator from Module 09 Lesson 3 - Claude re-reads brief.md, compares against recent session memories + plan.md, flags anything that's drifted or needs updating.

Action items

☐ Install the superpowers pack if you haven't (github.com/obra/superpowers) - gives you /brainstorming + /writing-plans ☐ For your next non-trivial project, start with a 10-question brainstorm

☐ Save the output as brief.md in the project folder

☐ Revisit the brief weekly - does it still reflect reality?

Next lesson: Writing a plan document.

Exercises

1. Review the concepts covered in this lesson: Brainstorm with Claude.
2. Write down your key takeaway from this lesson.
3. Practice running any commands or prompts mentioned above inside your terminal.