Brainstorming
Overview
Convert rough intent into a validated design with clear scope, constraints, and success criteria.
Why this exists
Skipping design causes the same failures repeatedly: wrong scope, wrong assumptions, and expensive rework. This skill keeps planning lightweight while still forcing the core decisions to be explicit.
Flow
- •Run a conversational clarification loop when request ambiguity is non-trivial.
- •Propose 2-3 approaches with practical trade-offs.
- •Recommend one approach and explain why.
- •Confirm design sections before implementation.
Conversational clarification loop
- •Ask exactly one question at a time.
- •Prefer multiple-choice questions where possible.
- •Keep the loop to 3-5 questions (max 5).
- •Focus questions on decisions that materially change implementation.
- •For low-ambiguity requests, ask 0-2 confirmation questions and proceed.
Stop the loop when either condition is met:
- •scope, constraints, and success criteria are explicit enough to plan safely
- •five questions have been asked
Before proposing approaches, emit a short decision log:
- •what changed from user answers
- •assumptions that remain
- •unresolved unknowns (if any)
Design sections to cover
- •Architecture shape
- •Components and responsibilities
- •Data flow and state boundaries
- •Error handling and fallback behavior
- •Validation strategy (manual by default; tests only when requested)
Output format
Return:
- •Recommended approach
- •Alternatives considered
- •Explicit scope in and out
- •Risks and assumptions
- •Decision log from conversational loop
Project Defaults
- •Keep planning lightweight and fast.
- •Do not force approval gates for each action.
- •Do not introduce worktrees unless user asks.
Red flags
- •Jumping into implementation without explicit scope
- •Presenting one option only
- •Treating constraints as optional
- •Asking a long list of questions in one message
- •Continuing clarification after scope is already explicit