Brainstorming

When to Use This Skill

Use brainstorming when:

• User mentions wanting to build/write something but hasn't structured it yet
• Initial idea is vague or broad ("I want to build notifications" or "I want to write about productivity")
• User is exploring multiple angles or approaches
• Idea needs refinement before planning or outlining

Skip when:

• User has clear requirements or outline ready
• Topic/project is well-defined and just needs execution
• User explicitly asks to skip ideation and start working

Important: If during brainstorming you realize the user is unclear about what they want (not just how to approach it), pause and use the interviewing skill first. Interviewing clarifies requirements and goals; brainstorming explores approaches and angles. Return to brainstorming once the what is clear.

Critical: User's Thoughts, Not Yours

Your role: Draw out the user's ideas through questions. Never inject your own ideas.

Use the AskUserQuestion tool to ask clarifying questions. This pauses execution and waits for the user's response, enabling true back-and-forth ideation.

ASK questions to explore their thoughts:

• "What triggered this topic?"
• "What's your core argument or insight?"
• "What examples from your experience illustrate this?"
• "Who is this for?"

DON'T suggest ideas they haven't mentioned:

❌ BAD (injecting):
AI: You should write about microservices vs monoliths

✓ GOOD (exploring with AskUserQuestion):
AI: [uses AskUserQuestion tool]
    What aspect of architecture are you thinking about?

The user is the expert on their own experience. You're just helping them structure it.

Core Approach

Start with Questions, Not Suggestions

Don't immediately propose outlines or structure. First understand:

• What triggered this topic? (specific experience, observation, frustration?)
• Who is the audience?
• What's the core insight or argument?
• What makes this topic relevant now?

Use AskUserQuestion for each clarifying question:

AI: [uses AskUserQuestion tool with question="What triggered this - specific experience or pattern you've noticed?"]

User: [responds with their thoughts]

AI: [uses AskUserQuestion tool with question="Is this for engineers, managers, or general audience?"]

Key questions to ask:

• "What triggered this - specific experience or pattern you've noticed?"
• "Is this for engineers, managers, or general audience?"
• "What's the contrarian take here? What does conventional wisdom miss?"
• "Why now? What makes this relevant or timely?"

Ideation Techniques

Use AskUserQuestion for all of these techniques - each question should pause and wait for user response.

1. Explore Tensions and Contradictions

Look for interesting conflicts:

• "You said X works, but also mentioned Y failed - what's the difference?"
• "That sounds like it contradicts Z - is that the point?"

2. Challenge Assumptions

Gently probe the premise:

• "Is that always true, or are there contexts where it breaks down?"
• "What would someone who disagrees with this say?"

3. Find the Concrete Angle

Move from abstract to specific:

• Vague: "I want to write about AI"
• Concrete: "Why AI code review misses context that human reviewers catch"

Pattern:

• Abstract topic → Specific problem
• General observation → Concrete example
• Theory → Practical implication

4. Suggest Multiple Perspectives

Offer 2-3 different angles, not just one:

• "You could approach this as: (1) why X fails, (2) what to do instead, or (3) when X actually works"
• "This could be prescriptive (here's how to fix it) or descriptive (here's why it happens)"

5. Use Personal Experience as Foundation

Ground abstract concepts:

• "You mentioned seeing this at 3 companies - what pattern did you notice?"
• "Walk me through a specific example where this happened"

Working with Outputs

For Writing (via /new-post):

Add ideas to braindump.md:

Structured sections:

• Context: What triggered this topic
• Core Argument: Main thesis or insight
• Audience: Who this is for
• Angles: Different approaches to explore
• Examples: Concrete instances, anecdotes
• Questions: Open questions to resolve

Iterate through conversation - update braindump.md as ideas evolve.

For Projects (via /orchestrate --discover):

Create discovery.md with:

Structured sections:

• Context: What triggered this project idea
• Problem Statement: What needs solving
• Constraints: Technical, time, resource constraints
• Approaches: Different technical approaches to consider
• Research Findings: Links, docs, patterns discovered
• Open Questions: Unclear requirements or unknowns

Transition Guidance

Know when to move from brainstorming to execution:

For Technical Projects

Ready to plan when:

• Problem statement is clear and specific
• Key constraints identified (performance, scale, security)
• 2-3 potential approaches explored
• User expresses readiness or no major unknowns remain

Transition:

AI: We've clarified:
    - Problem: Real-time notifications for 10k+ concurrent users
    - Constraints: Must integrate with existing auth, <100ms latency
    - Approach: WebSocket with Redis pub/sub
    - Open questions: Documented in discovery.md

    Ready to create implementation plan?
    → Recommend: /plan-feature [clarified request]
    → Next agent: planning-agent (uses technical-planning skill)

Not ready when:

• Problem statement is vague ("something with notifications")
• Critical constraints unknown (scale, security requirements)
• Multiple competing approaches without clarity

For Writing

Ready to outline when:

• Core argument is clear
• Audience is defined
• 2-3 concrete examples identified
• User expresses readiness ("okay, let's outline this")

Transition:

AI: We've got:
    - Core argument: OKRs fail because they measure what's easy, not what matters
    - 3 examples from your experience
    - Target audience: engineering managers

    Ready to structure this into an outline?
    → Continue in braindump.md with outline
    → Next: Draft in draft.md using blog-writing skill

Not ready when:

• Core argument is still fuzzy
• Multiple competing angles without clarity on which to pursue
• Missing concrete examples or evidence

Common Pitfalls to Avoid

1. Injecting Your Ideas: Don't suggest topics or angles the user hasn't mentioned - ask questions to draw out THEIR ideas
2. Premature Structuring: Don't jump to outline before the idea is clear
3. Too Many Options: Don't overwhelm with 10 different angles - offer 2-3
4. Leading the Witness: Ask genuine questions, don't push your preferred angle
5. Over-Abstracting: Keep pulling back to concrete examples
6. Ignoring Constraints: If user says "short post," don't brainstorm epic series
7. Making Up Examples: Don't invent scenarios - use only what the user has shared

Quality Checklist

Before transitioning (Writing):

• Core argument is clear and specific (not vague)
• At least 2-3 concrete examples or data points identified
• Audience and purpose are defined
• User feels ready to move forward
• Braindump.md has been updated with key ideas

Before transitioning (Projects):

• Problem statement is clear and specific (not vague)
• Key constraints identified (scale, performance, security, integrations)
• At least 2-3 potential approaches explored
• User feels ready to move forward or major unknowns documented
• Discovery.md has been created with findings and open questions

Example Flow

For detailed conversation examples showing brainstorming techniques in action, see reference/examples.md.

Integration with Other Skills

Before brainstorming (when needed):

• Use interviewing skill if the user is unclear about what they want - their goals, requirements, or purpose. Interviewing clarifies the "what"; brainstorming explores the "how".

For Writing:

• After brainstorming: Transition to blog-writing skill (writing plugin) for drafting
• During brainstorming: Use research-synthesis skill if research is needed
• Throughout: Update braindump.md with evolving ideas

For Projects:

• After brainstorming: Transition to technical-planning skill via planning-agent
• During brainstorming: Use research-synthesis skill to investigate approaches, docs, patterns
• Throughout: Update discovery.md with findings and clarifications