brainstorming

name	brainstorming
description	Use when starting design or discovery work. MUST use before any creative work.

Brainstorming

When to Use

∀ design ∧ discovery phases. Explores intent -> design -> approval.

HARD-GATE

Do NOT invoke any implementation skill, write any code, ∨ take any implementation action until design presented ∧ approved. Applies to EVERY project regardless of perceived simplicity. VIOLATION -> abort ∧ notify user. No exceptions, no shortcuts.

Conversation Rules

1. One question per message
2. Multiple choice preferred (A/B/C/D) — open-ended when needed
3. Assess scope first: multi-subsystem -> decompose before detailing
4. ∀ assumption: surface explicitly, ¬proceed on implicit agreement
5. Propose 2-3 approaches w/ trade-offs before committing
6. ∃ question answerable by codebase exploration -> explore, ¬ ask
7. APPROACH ∧ DESIGN questions: ship recommended answer (user confirms ∨ overrides). FRAME questions: open, ¬ recommendation (avoid biasing problem framing)
8. Walk decision-tree: resolve dependent decisions before independent ones; ∀ branch -> exhaust before next

Flow

1. FRAME: Define problem, constraints, scope (2-4 open questions, ¬ recommendations)
2. APPROACH: Propose 2-3, recommend one, user picks
3. DESIGN: Branch-traversal — ∀ decision: state question + recommended answer + dependencies; user confirms ∨ overrides; resolve children before siblings
   • standard: problem, approach, acceptance criteria, non-goals (~1 page)
   • complex: + constraints, architecture, error handling, testing strategy (~3 pages)
4. WRITE: project spec document (e.g., docs/specs/SPEC.md)
5. REVIEW: Dispatch anonymous spec reviewer subagent (max 3 iterations)
6. USER GATE: Show spec, ask approval inline

Spec Document Reviewer

Dispatch via Agent tool (subagent_type: general-purpose) with prompt from brainstorming review template. Max 3 iterations.

Calibration: Only flag issues that cause real problems during planning. Not theoretical concerns.

Escalation: 3+ failed iterations -> abort review, create escalation task for human review. Never infinite retry loops.

Plan Document Reviewer

Dispatch via Agent tool (subagent_type: general-purpose). Checks: completeness, task decomposition, buildability, TDD, traceability, YAGNI, file structure. Max 3 iterations.

Escalation: 3+ failed iterations -> abort review, create escalation task for human review.

Anti-Patterns

• "This is too simple to need a design" — WRONG. Simple projects are where unexamined assumptions cause most wasted work.
• Implementing before design approved
• Skipping the FRAME phase (defining problem before solving)
• Rubber-stamping spec review iterations
• Asking the user something a grep ∨ read could resolve
• Recommending answers in FRAME phase (biases problem definition)
• Linear question dump, ¬ following dependency edges of the decision-tree

Rules

• ∀ section: user approves before next
• ∀ spec: reviewed by anonymous subagent
• ∀ approach: trade-offs documented
• ¬implement anything until design approved