---

name: brainstorming description: Fast ideation and problem exploration. Triggers on Ukrainian/English phrases like "подумай", "поміркуй", "брейнштормінг", "brainstorm", "think through", "накидай ідеї", "generate ideas". Rapid divergent thinking to uncover insights, problems, and solutions through interactive dialogue.

Brainstorming

Overview

Quick, interactive ideation for exploring problems and generating solutions. Focus on speed, quantity, and dialogue - not lengthy analysis.

When to Use

• User asks: "подумай", "поміркуй", "brainstorm this"
• Need to explore problem space rapidly
• Generate multiple ideas quickly
• Uncover hidden insights or issues
• User explicitly requests brainstorming

Core Principles

1. Quantity over quality - Generate many ideas fast
2. No judgment during ideation - Wild ideas welcome
3. Interactive - Communicate with user frequently
4. Iterative - Build on ideas through dialogue
5. Fast - Minutes, not hours

Workflow

Step 1: Rapid Dump (1-2 minutes)

Immediately generate 5-10 initial ideas without overthinking:

CopyQuick ideas:
• [Idea 1]
• [Idea 2]
• [Idea 3]
• [Idea 4]
• [Idea 5]
...

Rules:

• Write first idea that comes to mind
• No filtering or judgment
• Include obvious AND wild ideas
• One line per idea

Step 2: Communicate & Explore

Share ideas with user and ask:

• Which direction interests you?
• What concerns do you have?
• What am I missing?
• Should we explore deeper or pivot?

Keep it conversational - don't wait to have "perfect" analysis.

Step 3: Iterate Based on Feedback

Depending on user response:

• Expand promising direction - Generate more variations
• Find problems - What could break? What's risky?
• Challenge assumptions - What if we did opposite?
• Combine ideas - Mix two approaches

Step 4: Quick Synthesis (when needed)

If user asks for recommendation, provide:

CopyTop pick: [Option X]
Why: [1-2 sentence reason]
Next step: [Immediate action]

Keep it short - avoid long justifications unless requested.

Techniques

Random Prompts

• "What if we did the opposite?"
• "How would [different domain] solve this?"
• "What's the simplest version?"
• "What's the most expensive/complex version?"

SCAMPER (quick version)

• Substitute - swap components
• Combine - merge ideas
• Adapt - borrow from elsewhere
• Modify - change scale/attribute
• Put to other use - repurpose
• Eliminate - remove parts
• Reverse - flip it around

5 Whys (fast)

Ask "why?" 3-5 times to find root issue.

Perspective Shifts

• User perspective
• Developer perspective
• Future-self perspective

Output Style

Good:

CopyIdeas:
• Use caching
• Pre-compute results
• Lazy load data

Which feels right? Or should we explore something else?

Bad:

Copy# Comprehensive Analysis of Caching Strategies

After extensive research into caching mechanisms, we can evaluate several approaches...

[5 more paragraphs]

Anti-Patterns

• ❌ Writing long analysis before sharing ideas
• ❌ Waiting for "complete" answer before engaging user
• ❌ Judging ideas during generation phase
• ❌ Fixating on first solution
• ❌ Over-explaining instead of asking

Example Session

User: "подумай як зробити швидший пошук"

Skill:

CopyIdeas for faster search:
• Add database indexes
• Cache popular queries
• Use full-text search engine (Elasticsearch)
• Pre-compute search results
• Reduce search scope with filters
• Client-side filtering for small datasets

Which direction makes sense for your use case?
How much data are we talking about?

User: "Maybe 10K items, need instant results"

Skill:

CopyFor 10K items, instant results:

Top 3 approaches:
• Client-side search (send all data, filter in browser) - simplest
• Add indexes + optimize query - medium effort
• Redis cache for popular searches - fastest for repeated queries

Concerns:
• 10K items might be too much for client-side if complex objects
• Need to know query patterns

What kind of queries? Full-text? Exact match? Filters?

Notes

• Brainstorming is dialogue, not monologue
• Share incomplete thoughts - perfection kills speed
• Ask questions frequently
• User steers direction, not you
• Stop when user has enough to proceed