Brainstorming with Cognitive Lenses

Version: 1.0.365+

/sw:brainstorm is a structured ideation skill that explores problems from multiple perspectives before you commit to an implementation path. It uses established cognitive frameworks (Six Thinking Hats, SCAMPER, TRIZ) to expand the solution space beyond the "obvious" answer.

Why This Exists

When you ask an AI to brainstorm, you typically get 3-4 approaches ranked by popularity — JWT first for auth, React first for frontends, PostgreSQL first for databases. The same list every time.

Cognitive lenses force exploration off the beaten path. The Black Hat finds risks everyone ignores. SCAMPER's "Eliminate" asks what you can remove entirely. TRIZ inverts your core assumptions. These aren't random — they're backed by decades of creativity research.

The Gap It Fills

Without brainstorming, the path from "idea" to "spec" is a straight line — and that's where most bad architectural decisions get locked in. /sw:brainstorm adds a structured exploration phase that saves hours of rework later.

Quick Start

# Standard: 4-6 approaches with comparison matrix
/sw:brainstorm "how should we handle real-time notifications"

# Quick: 3 approaches, fast comparison, 2 minutes
/sw:brainstorm "JWT vs OAuth vs Passkeys" --depth quick

# Deep: multiple cognitive lenses, full analysis
/sw:brainstorm "microservices vs monolith for our SaaS" --depth deep

# Specific lens
/sw:brainstorm "payment gateway architecture" --lens six-hats

# Resume a previous session
/sw:brainstorm "payment gateway" --resume

# Custom evaluation criteria
/sw:brainstorm "marketing strategy" --criteria "brand-fit,audience-reach,cost,differentiation"

The 5-Phase Flow

Phase 1: Frame

Every brainstorm starts by sharpening the problem. Most bad solutions come from solving the wrong problem.

What happens:

• Restates your topic as a clear, one-sentence problem statement
• Runs Starbursting (5W1H): Who/What/When/Where/Why/How
• Asks 1-2 targeted clarifying questions

Example: You say "we need caching." Starbursting reveals the real pain is N+1 database queries, not cache misses. The brainstorm pivots from "which cache" to "how to eliminate redundant queries" — a fundamentally different and better question.

Phase 2: Diverge

This is the core innovation. Instead of generating approaches from one perspective, cognitive lenses force the model to think from structurally different angles.

5 available lenses:

Lens	Perspectives	Best For
Default	Conservative, Bold, Speed, Extensibility, Lateral, Hybrid	General engineering decisions
Six Thinking Hats	Facts, Feelings, Caution, Optimism, Creativity, Process	Team decisions, product strategy
SCAMPER	Substitute, Combine, Adapt, Modify, Repurpose, Eliminate, Reverse	Improving existing systems
TRIZ	13 inventive principles + constraint inversion	Breaking through technical contradictions
Adjacent Possible	What recently became feasible (web-search-enhanced)	New products, tech modernization

In deep mode, each lens facet is dispatched as a separate parallel subagent — Six Thinking Hats literally runs 6 agents simultaneously, each thinking from one perspective.

Phase 3: Evaluate

All approaches are scored on a comparison matrix (1-5 scale):

| Criterion     | Approach A | Approach B | Approach C |
|---------------|:----------:|:----------:|:----------:|
| Complexity    |    2/5     |    4/5     |    3/5     |
| Time          |    4/5     |    2/5     |    3/5     |
| Risk          |    4/5     |    3/5     |    4/5     |
| Extensibility |    2/5     |    5/5     |    3/5     |
| Alignment     |    5/5     |    2/5     |    4/5     |
| Total         |    17      |    16      |    17      |

Criteria are customizable — use --criteria for domain-specific evaluation (marketing, infrastructure, business).

An explicit recommendation with rationale is provided. You can accept, explore deeper, or run more lenses.

Phase 4: Deepen (Deep Mode Only)

Goes beyond comparison into structural analysis:

• Abstraction Laddering: Zoom out (are we solving the right problem?) and zoom in (what are the first 3 concrete steps?)
• Analogical Reasoning: "This is similar to how Netflix solved X" — distant-field analogies that reveal non-obvious solutions
• Hidden Assumptions: List implicit assumptions and invert each one to check for blind spots
• Pre-Mortem: "Imagine this approach failed. Why?" — Gary Klein's technique for catching risks that optimistic planning misses

Phase 5: Output

Saves a persistent brainstorm document to .specweave/docs/brainstorms/YYYY-MM-DD-topic.md and offers handoff:

1. Turn into increment — passes brainstorm context (problem frame, selected approach, risks, constraints) to /sw:increment so the PM agent has a head start
2. Go deeper — re-run with different lenses or deeper mode
3. Park for later — the document and idea tree persist for revisiting

Depth Modes

Mode	Phases	Duration	Use When
quick	1 + 3	~2 min	Binary decisions, simple comparisons
standard	1-3 + 5	~5 min	Typical feature decisions, architecture choices
deep	All 5	~15 min	Major architectural decisions, new product ideation, high-stakes choices

Cognitive Lenses in Detail

Default Lens

Generates 4-6 approaches with different strategic orientations:

1. Conservative: Build on what exists. Minimal change, maximum reuse.
2. Bold: Rethink from scratch. What's ideal with no constraints?
3. Speed: Optimize for fastest delivery. Simplest thing that works.
4. Extensibility: Optimize for future growth. What won't you regret in 2 years?
5. Lateral (optional): What would a completely different industry do?
6. Hybrid (optional): Combine the best parts of other approaches.

Six Thinking Hats (de Bono)

Forces 6 structurally different perspectives:

Hat	Focus	Example Output
White (Facts)	Data, evidence, metrics	"Our current auth handles 10K req/s. JWT validation adds 2ms latency."
Red (Feelings)	Intuition, user emotions	"Users hate password reset flows. Passkeys feel magical."
Black (Caution)	Risks, worst cases	"If the OAuth provider goes down, all users are locked out."
Yellow (Optimism)	Opportunities, best cases	"Passkeys unlock biometric login — 3x faster onboarding."
Green (Creativity)	Novel, unconventional	"What if auth is optional? Public-by-default with audit trails."
Blue (Process)	Structure, methodology	"Phase 1: JWT for MVP. Phase 2: Add OAuth. Phase 3: Passkeys for power users."

When to use: Product decisions with multiple stakeholders. Marketing strategy. Any decision where emotional and political factors matter as much as technical ones.

SCAMPER

7 systematic transformations applied to the current system:

Letter	Question	Example
Substitute	What can we replace?	Replace REST with GraphQL
Combine	What can we merge?	Combine auth + billing into one service
Adapt	What can we borrow?	Adapt Stripe's hosted checkout pattern
Modify	What can we scale up/down?	Minimize the API surface to 5 endpoints
Put to other use	What can we repurpose?	Use the audit log as an analytics source
Eliminate	What can we remove?	Eliminate the custom admin panel — use Retool
Reverse	What can we invert?	Push notifications instead of pull/polling

When to use: Improving or rethinking existing systems. Reducing complexity. Finding optimizations.

TRIZ (Inventive Principles + Constraint Inversion)

Applies selected inventive principles from Altshuller's 40 TRIZ principles, adapted for software:

Principle	Software Adaptation
Segmentation	Break monolith into microservices
Extraction	Extract cross-cutting concerns into middleware
Merging	Combine multiple API calls into batch endpoints
Preliminary Action	Pre-compute, cache, warm up at build time
The Other Way Round	Invert control flow (push vs pull, event sourcing)
Dynamicity	Feature flags, config-driven behavior
Blessing in Disguise	Turn constraints into features (rate limiting becomes fair usage)
Intermediary	Add proxy, gateway, anti-corruption layer
Self-Service	User-facing admin panels, API key management

Then inverts core assumptions: "Users must authenticate first" → "What if data is public by default?" This is how Google Docs sharing works.

When to use: Technical contradictions. "We need X but it conflicts with Y." Breaking through seemingly impossible constraints.

Adjacent Possible

Web-search-enhanced lens that asks: "What recently became feasible?"

Uses live web search to ground ideas in reality — not just model knowledge but actual recent developments, new tools, cost changes, and emerging patterns.

When to use: New products. Modernization efforts. "Should we build or buy?" decisions. Any situation where the landscape changed recently.

Resume Mode

Brainstorms are persistent. Every session saves a state file and document that you can resume later:

# Start a brainstorm
/sw:brainstorm "notification architecture"

# ... come back later and go deeper
/sw:brainstorm "notification architecture" --resume --depth deep --lens triz

Resume mode:

• Loads the previous session's problem frame and approaches
• Skips phases you already completed
• Lets you explore abandoned branches from the idea tree
• Applies new lenses to the same problem with accumulated context

Custom Evaluation Criteria

The default criteria (Complexity, Time, Risk, Extensibility, Alignment) work for engineering decisions. For other domains, use --criteria:

# Marketing brainstorm
/sw:brainstorm "launch strategy" --criteria "brand-fit,audience-reach,cost,differentiation"

# Database selection
/sw:brainstorm "database choice" --criteria "read-perf,write-perf,ops-complexity,cost,ecosystem"

# Business decision
/sw:brainstorm "pricing model" --criteria "revenue-impact,customer-retention,implementation-effort,competitive-position"

Preset auto-detection: If you don't specify --criteria, the skill detects context and picks appropriate criteria:

Context	Auto-Detected Criteria
Engineering	Complexity, Time, Risk, Extensibility, Alignment
Marketing/Product	Brand Fit, Audience Reach, Cost, Differentiation, Time-to-Market
Infrastructure	Performance, Reliability, Cost, Operational Complexity, Scalability
Business	Revenue Impact, Cost, Time-to-Value, Strategic Alignment, Risk

Real-World Examples

Example 1: Auth Architecture Decision

/sw:brainstorm "authentication for our B2B SaaS" --depth deep --lens six-hats

What you get:

• White Hat: "Current competitors use OAuth2 + SAML. Enterprise customers require SSO."
• Red Hat: "Developers hate SAML integration. Users love magic links."
• Black Hat: "Custom auth is the #1 source of security vulnerabilities."
• Yellow Hat: "Passkeys eliminate password-related support tickets entirely."
• Green Hat: "What if we delegate 100% of auth to a managed provider and own zero credentials?"
• Blue Hat: "Phase 1: Auth0 for MVP. Phase 2: Add SAML for enterprise. Phase 3: Passkeys for end users."

Evaluation matrix scores each on Complexity/Time/Risk/Extensibility/Alignment. Recommendation: Blue Hat's phased approach — lowest risk, fastest start, clear upgrade path.

Example 2: Payment Flow Optimization

/sw:brainstorm "our checkout flow has 40% drop-off" --lens scamper

SCAMPER produces:

• Substitute: Replace multi-step form with single-page checkout
• Combine: Merge shipping + billing into one address entry
• Eliminate: Remove account creation requirement (guest checkout)
• Reverse: Show order summary first, then collect payment (Amazon's approach)
• Adapt: Borrow Apple Pay's one-click pattern

Example 3: New Product Ideation

/sw:brainstorm "developer productivity tool for code review" --depth deep --lens adjacent

Adjacent Possible uses web search to find:

• AI code review tools that launched in 2025-2026
• Cost of LLM API calls dropping below $1/1M tokens
• New IDE extension APIs enabling inline suggestions
• Emerging patterns from CodeRabbit, Graphite, etc.

Generates approaches grounded in what's actually possible today, not generic "use AI" suggestions.

Integration with SpecWeave Workflow

The brainstorm output feeds directly into the increment workflow:

• Problem frame → PM agent skips basic discovery questions
• Selected approach → Architect agent starts with a direction instead of from scratch
• Risks and constraints → get converted into acceptance criteria
• Evaluation matrix → becomes a reference for architecture decision records (ADRs)

The brainstorm document persists at .specweave/docs/brainstorms/ and is bidirectionally linked to its resulting increment.

Output Location

Artifact	Path
Brainstorm document	.specweave/docs/brainstorms/YYYY-MM-DD-topic-slug.md
Session state	.specweave/state/brainstorm-YYYY-MM-DD-topic-slug.json

Tips

• Start quick, go deep later: Use --depth quick for initial exploration, then --resume --depth deep on the most promising direction
• Combine lenses: In deep mode, select multiple lenses to get perspectives from different frameworks
• Use for marketing too: Six Thinking Hats works great for positioning, messaging, and go-to-market decisions — not just engineering
• Share the document: The brainstorm doc is markdown — paste it in a PR description, Slack thread, or design review for team alignment
• Revisit abandoned branches: Use --resume to explore approaches you initially discarded — sometimes they become viable as constraints change

Related

• Planning Workflow — The full planning flow that brainstorming precedes
• Skills Reference: brainstorm — Quick reference for all arguments
• Deep Interview Mode — Complementary feature for detailed requirement gathering