AgentSkillsCN

project-brainstorm

将零散的创意、灵感与待解难题转化为结构化的项目概念,为后续规格制定做好准备。

SKILL.md
--- frontmatter
name: project-brainstorm
description: Transform loose ideas, inspirations, and problems into structured project concepts ready for specification
disable-model-invocation: false

description: Transform loose ideas, inspirations, and problems into structured project concepts ready for specification.

User input:

$ARGUMENTS

The text the user typed after /project-brainstorm in the triggering message. This may be:

  • Stream of consciousness thoughts
  • Problem statements
  • Inspirations from other products
  • "I wish I had..." statements
  • Vague concepts
  • Multiple unrelated ideas
  • Or nothing (start interactive session)

Purpose

This command helps users who have ideas but haven't crystallized them into a specific project yet. It bridges the gap between "I have this vague idea" and "I'm ready to specify a project."

Position in Flow: BEFORE /project-specify (optional ideation phase)

Output: One or more project concepts ready for /project-specify

Brainstorm Process

Step 1: Idea Gathering

If user provided ideas in $ARGUMENTS:

  • Parse and acknowledge each distinct idea/problem/inspiration
  • Ask: "Is there anything else you'd like to add before we explore these?"

If no arguments provided, start interactive session:

code
💡 Let's brainstorm! Tell me about:
- Problems you want to solve
- Things that frustrate you
- Products you wish existed
- Features you want to build
- Industries or domains you're interested in
- Technologies you want to explore

Just brain dump - no structure needed. I'll help organize.

Continue gathering until user signals they're done or you have enough material.

Step 2: Theme Clustering

Analyze the ideas and identify themes:

Problem Themes:

  • What problems keep appearing?
  • Who has these problems?
  • How severe are they?

Solution Themes:

  • What capabilities would address these?
  • What technologies could help?
  • What existing solutions are inadequate?

User Themes:

  • Who are the potential users?
  • What are their contexts?
  • What motivates them?

Domain Themes:

  • What industries or fields?
  • What expertise is relevant?
  • What regulations apply?

Present clustering:

code
📊 I see these themes in your ideas:

**Problems**:
1. [Theme 1] - Examples: [ideas that fit]
2. [Theme 2] - Examples: [ideas that fit]

**Potential Users**:
- [User type 1]: [their context]
- [User type 2]: [their context]

**Domain/Industry**:
- [Domain] with [specific focus]

Does this capture the essence? Anything to add or correct?

Step 3: Concept Generation

For each viable theme combination, generate a project concept:

Concept Template:

code
📋 Project Concept: [Name]

**One-Liner**: [Single sentence describing the product]

**Problem Statement**: 
[Who] experiences [what problem] when [context], which causes [impact].

**Proposed Solution**:
A [product type] that [core capability], enabling users to [key benefit].

**Target Users**:
- Primary: [User type with context]
- Secondary: [Other user types]

**Key Features** (high-level):
1. [Feature 1]
2. [Feature 2]
3. [Feature 3]

**Differentiation**:
Why this is better than [existing solutions]: [reason]

**Complexity Estimate**:
- Scale: Level [0-4] (atomic → platform)
- Timeframe: [Rough estimate]
- Technical: [Key technical challenges]

**Confidence**: [High/Medium/Low] - [Why]

Generate 1-3 concepts depending on idea diversity.

Step 4: Concept Exploration

For each concept, offer to explore further:

code
Would you like me to:
1. 🔍 Deep dive on "[Concept Name]" - explore feasibility, competition, risks
2. 🔀 Combine concepts - merge [A] and [B] into something new
3. ➕ Generate more variations - different angles on the same problem
4. ✅ Proceed with "[Concept Name]" to /project-specify
5. 💭 Keep brainstorming - add more ideas to the mix

Step 5: Concept Refinement (if requested)

Deep Dive includes:

  • Quick web search for existing solutions
  • Identify potential competitors
  • Technical feasibility assessment
  • Risk identification
  • Market opportunity (if applicable)

Combination includes:

  • Merge complementary features
  • Identify synergies
  • Assess increased complexity
  • Generate new unified concept

Step 6: Output & Handoff

When user selects a concept to proceed with:

  1. Generate project description suitable for /project-specify:
code
📝 Ready for Specification!

Here's your project description for /project-specify:

---
[Refined one-paragraph description that captures:
- What it is
- Who it's for
- Core problem solved
- Key differentiator]
---

Run this to continue:
/project-specify [paste description above]

Or I can run it now with this description. Proceed?
  1. Save brainstorm session (optional):
code
Would you like me to save this brainstorm session?
→ Saves to specs/projects/brainstorm-[timestamp].md
→ Useful for revisiting rejected concepts later

Brainstorm Artifacts

If saving, create:

markdown
# Brainstorm Session: [Date]

## Raw Ideas
[User's original input]

## Themes Identified
### Problems
- [List]

### Users
- [List]

### Domains
- [List]

## Concepts Generated

### Concept 1: [Name]
[Full concept template]

### Concept 2: [Name]
[Full concept template]

## Selected Concept
[Name] - Proceeding to /project-specify

## Rejected Concepts (for future reference)
- [Concept]: [Why not selected]

Integration with /speck

The /speck router should detect brainstorm intent:

  • "I have an idea but it's not clear yet"
  • "Help me figure out what to build"
  • "I want to brainstorm"
  • "Let's explore some ideas"
  • Vague descriptions that lack specificity

Route to /project-brainstorm instead of /project-specify.

Examples

Example 1: Vague Idea

code
User: /project-brainstorm I want to build something with AI that helps people

Agent: 💡 That's a great starting point! Let's explore...
       - What kind of help? (productivity, health, learning, creativity?)
       - Which people? (students, professionals, elderly, general public?)
       - What frustrates you about current AI tools?
       
[Continues interactive exploration]

Example 2: Multiple Ideas

code
User: /project-brainstorm I hate how hard it is to find good restaurants,
      and also I want to track my reading habits, and maybe something 
      with local events

Agent: 📊 I see three distinct areas in your ideas:
       1. **Discovery Problem**: Finding quality restaurants
       2. **Personal Tracking**: Reading habit management  
       3. **Local Awareness**: Event discovery
       
       These could be separate projects OR combined into a 
       "local lifestyle" platform. Let me generate concepts...
       
[Generates concepts for each and combined]

Example 3: Problem Statement

code
User: /project-brainstorm It's so frustrating when I want to hang out with 
      friends but coordinating schedules is a nightmare

Agent: 🎯 Clear problem! Let me explore the space:
       
       **Problem**: Schedule coordination for social activities
       **Who**: Friend groups trying to meet up
       **Pain Points**: 
       - Back-and-forth messaging
       - Finding times that work for everyone
       - Picking activities/locations
       
       Let me generate concepts that address this...
       
[Generates 2-3 concepts: simple scheduler, social planning app, 
availability matching platform]

Tips for Good Brainstorming

Do:

  • Ask clarifying questions to understand intent
  • Explore tangential ideas - they often lead somewhere
  • Challenge assumptions about what the user "really" needs
  • Suggest combinations and variations
  • Be encouraging about half-formed ideas

Don't:

  • Rush to specification too quickly
  • Dismiss ideas as "too simple" or "already done"
  • Force structure before exploration is complete
  • Ignore the emotional/personal motivation behind ideas
  • Forget to save interesting rejected concepts

Research Integration

Use web search to:

  • Find existing solutions (competitors)
  • Validate that problem exists
  • Identify market gaps
  • Find technology options

Document findings in concept generation:

code
**Quick Research Findings**:
- Existing solutions: [List with limitations]
- Market gap: [What's missing]
- Technology: [What enables this now]

Success Criteria

A successful brainstorm session:

  • ✅ User feels heard and understood
  • ✅ At least one concrete project concept generated
  • ✅ Concept is specific enough for /project-specify
  • ✅ User is excited about the direction
  • ✅ Rejected concepts are preserved for future reference