Brainstorming — Reframe Before You Build

SkillMedia

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Reframes the problem, challenges assumptions, and outputs a design doc that downstream skills can read.

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the Brainstorming skill

What this skill tells your AI

The instructions your AI receives, as published by myths-labs/muse in skills/toolkit/brainstorming/SKILL.md and read by ahel’s review.

Inspired by gstack's /office-hours. Upgraded to output standardized design docs.

Overview

Don't just ask "what do you want to build?" Ask "what pain are you solving?" This skill reframes the problem before a line of code is written, then outputs a design doc that feeds into /sprint downstream skills.

The Process

Phase 1: Reframe the Problem

Before exploring solutions, challenge the framing:

  1. Listen for pain, not features. User says "add a dark mode toggle." Ask: "What's the actual problem? Is it eye strain? Brand consistency? User requests?"
  2. Push back on scope. User describes 10 features. Ask: "If you could only ship one this week, which one?"
  3. Challenge 3 premises. Identify the user's hidden assumptions and question them explicitly.
    • "You're assuming users want X. What if they actually want Y?"
    • "You're assuming this needs to be real-time. Does it?"
    • "You're assuming mobile-first. What if desktop is the real use case?"

Rules:

  • One question per message
  • Multiple choice when possible
  • Don't rush past reframing — this is the highest-leverage phase

Phase 2: Explore Approaches

Once the real problem is clear:

  1. Generate 3 approaches with different tradeoffs:
    • Narrowest wedge — ship in hours, validate the thesis
    • Balanced — ship in days, covers core use cases
    • Full vision — ship in weeks, comprehensive solution
  2. Recommend one with clear reasoning
  3. Estimate effort for each (hours, not story points)

Lead with your recommendation. Explain why.

Phase 3: Present the Design

Once approach is chosen:

  • Break into sections of 200-300 words
  • After each section, ask: "Does this look right so far?"
  • Cover: architecture, components, data flow, error handling, testing
  • Apply ETHOS.md: Boil the Lake (do the complete thing), YAGNI (don't add what's not needed)

Phase 4: Output Design Doc

Write a standardized design doc to docs/plans/YYYY-MM-DD-<topic>-design.md:

# [Feature Name] Design Doc

## Problem
[1-2 sentences: the actual pain, not the feature request]

## Reframe
[How the problem was reframed during brainstorming]

## Approach
[Chosen approach and why. Brief mention of alternatives considered.]

## Scope
[What's IN scope and what's explicitly OUT]

## Technical Design
[Architecture, components, data flow, key decisions]

## Acceptance Criteria
[Numbered list of verifiable outcomes]

## Test Strategy
[What to test and how]

## Risks
[Top 2-3 risks and mitigations]

This doc is read by downstream skills:

  • writing-plans reads it to create implementation plan
  • architect-agent reads it for architecture review
  • code-reviewer-agent reads Acceptance Criteria to verify completeness

Key Principles

  • Reframe first — The best feature is often not what was originally requested
  • One question at a time — Don't overwhelm with multiple questions
  • Ship the Narrowest Wedge (from ETHOS.md) — Scope tight, validate fast
  • Three premises — Always challenge at least 3 hidden assumptions
  • Design doc is the handoff — Other skills read it, so make it precise
  • Be flexible — Go back and clarify when something doesn't make sense

Signals

GitHub stars
35
Forks
4
Last commit
Sep 2026
Hacker News mentions
14
Advanced
Item type
skill
Key
brainstorming-myths-labs
Source
github.com/myths-labs/muse