Brainstorming Requirement Clarity

Problem: Clarify WHAT Before Deciding HOW

When requests include vague phrasing such as 'make it better' or 'add something like X', implementation can quickly optimize the wrong goal. This skill is for AI agents to clarify WHAT to build before coding: purpose, users, constraints, success criteria, and boundaries. It is useful when requirements are ambiguous, multiple approaches are viable, or trade-offs need discussion with the user. It can be skipped when acceptance criteria, reference patterns, and scope are already explicit, or when the task is a straightforward bug fix.

How It Works: One-Question Clarification, Option Trade-offs, and Captured Decisions

The skill runs in phases:
- Phase 0: assess requirement clarity first, avoiding unnecessary questioning on already clear tasks.
- Phase 1: ask one question at a time, prefer multiple-choice options when natural, narrow from core purpose to users and constraints, and state assumptions explicitly.
- Phase 2: propose 2-3 concrete approaches with a recommendation, trade-offs, and YAGNI-driven simplicity, referencing existing codebase patterns when relevant.
- Phase 3: summarize key decisions into a short design record, suggested to be stored as docs/brainstorms/YYYY-MM-DD--brainstorm.md.
- Phase 4: let the user choose to proceed to planning, refine further, or pause.

It keeps each section short and pauses for validation after key outputs, reducing the risk of elaborating on a misaligned design.

Fit and Limits

This skill is a front-of-process clarification aid, not an implementation planner. It fits feature scoping, ambiguous product requirements, and unresolved architecture trade-offs. It may be redundant for clear, single-path tasks. Avoid asking many questions at once, jumping into implementation details, ignoring existing patterns, unvalidated assumptions, or producing long design documents instead of concise decisions.