Use this skill proactively when the user wants to explore ideas before committing. Trigger on: - Conversational: "let's brainstorm", "let me think", "hmm", "thinking about" - Ideation: "brainstorm", "what if", "ideas for", "possibilities", "spitball" - Exploration: "explore options", "think about", "how could we", "what are the options" - Uncertainty: "not sure where to start", "undecided", "weighing options", "torn between" - Design thinking: "alternative approaches", "creative solutions", "different ways" - Open-ended: "how might we", "what would it look like", "ways to" Provides structured brainstorming without judgment—all ideas are valid during divergent thinking.
This skill inherits all available tools. When active, it can use any tool Claude has access to.
A framework for exploring possibilities before committing to an approach.
Effective brainstorming separates two distinct phases:
DIVERGENT ──────────────▶ CONVERGENT
Generate ideas Evaluate & narrow
Quantity over quality Quality over quantity
No judgment Apply criteria
Wild is welcome Practical focus
Critical: Never mix these modes. Judging during divergent thinking kills creativity.
1. Direct Brainstorm List as many approaches as possible without filtering:
2. Analogies How do similar problems get solved elsewhere?
3. Constraints Inversion What if we changed the constraints?
4. Stakeholder Perspectives What would different people prioritize?
After generating options, apply criteria:
| Option | Effort | Risk | Maintainability | Fits Patterns |
|---|---|---|---|---|
| A | Low | Low | High | Yes |
| B | High | Low | High | Partial |
| C | Low | High | Medium | No |
Consider each option against:
Structure brainstorming output as:
## Brainstorm: [Topic]
### Context
[What problem or decision we're exploring]
### Ideas Generated
1. **[Idea Name]** - Brief description
2. **[Idea Name]** - Brief description
...
### Evaluation
| Idea | Pros | Cons | Effort |
|------|------|------|--------|
| 1 | ... | ... | Low |
| 2 | ... | ... | High |
### Recommended Direction
[Which 1-2 options to explore further and why]
### Open Questions
[What we'd need to answer to decide]
| Situation | Use |
|---|---|
| Unknown codebase | Research first (/harness:research) |
| Known code, unclear approach | Brainstorm |
| Multiple valid solutions | Brainstorm |
| Need to understand what exists | Research |
| Need to explore what's possible | Brainstorm |
After brainstorming, the path depends on what you converged on:
If the outcome is a multi-feature idea (game, app, system):
→ /harness:research to discover milestones
Research will break the idea into 3-7 plannable milestones. Then plan one milestone at a time.
If the outcome is one bounded feature:
→ /harness:plan "feature name" directly
If more ideation needed: → Continue brainstorming or revisit constraints
Before suggesting next steps, evaluate the brainstorm output:
| Output Type | Example | Next Step |
|---|---|---|
| Project | "Build a colony sim game" | /harness:research |
| MVP/PoC | "Minimal playable prototype" | /harness:research |
| Feature | "Add meeting room booking" | /harness:plan |
| Component | "Create the mood system" | /harness:plan |
Rule: If it would take more than 15 tasks to build, it needs research first.
/harness:research - Discover milestones from project ideas/harness:plan - Plan one milestone (10-15 tasks max)/milestone-planning - Guidance on breaking large work into phases