creative

name	creative
description	Divergent ideation and creative brainstorming. Use when you need novel solutions, want to break out of conventional thinking, or need multiple distinct approaches.
argument-hint	[problem or challenge to brainstorm]
user-invocable	true
allowed-tools	Read, Grep, Glob, Bash

Generate diverse, creative solutions for $ARGUMENTS using structured divergent thinking. Follow every step — creativity without structure produces noise, not options.

Step 1 — Understand before ideating

Before generating a single idea, understand the problem deeply:

Problem as stated: [what the user asked for]
Who has this problem: [specific people in specific situations]
Why it matters: [consequences of not solving it]
What's been tried: [existing approaches and why they're insufficient]
Constraints: [real constraints — budget, time, technical, organisational]

Do not skip this step. The most common failure mode in brainstorming is solving the wrong problem creatively.

Step 2 — Reframe the problem (3 ways)

The way you frame a problem determines the solutions you can see. Reframe the problem in three fundamentally different ways:

Reframe 1 — As a user story

"A [specific person] who needs [outcome] because [motivation], but currently [obstacle]."

This frame centres the human and their context. It often reveals that the real problem is different from the stated problem.

Reframe 2 — As a constraint

"We can't do [assumed approach], so what else could work?"

Remove the most obvious solution and see what emerges. This forces you past the first idea.

Reframe 3 — As an analogy

"This is like [something from a completely different domain] because [shared structure]."

Cross-domain analogies break fixedness. How does nature solve this? How does a restaurant solve this? How does a game solve this?

After all three reframes, pick the framing that opens the most creative space. State which one and why. This becomes the basis for ideation.

Step 3 — Generate options using diversity-forcing techniques

Produce at least 5 genuinely different approaches. "Genuinely different" means each option represents a fundamentally different strategy, not a variation on the same idea.

Mandatory diversity techniques

Apply each technique to generate at least one option. Not every technique will produce a winner — that's fine. The goal is to ensure you've explored the solution space widely.

Technique 1 — Inversion

What if you did the opposite of the obvious approach?

• If the obvious solution adds something, what if you removed something instead?
• If the obvious solution automates, what if you made it manual but better?
• If the obvious solution centralises, what if you distributed?

Generate one option from inversion.

Technique 2 — Extreme scale

What if this needed to serve 1 user? What about 1 million?

• The 1-user version strips away infrastructure and reveals the core value
• The 1-million-user version forces architectural thinking and reveals bottlenecks
• Often the best solution lives between these extremes but is visible from one of them

Generate one option from extreme scale thinking.

Technique 3 — Remove constraints

What if [biggest constraint] didn't exist?

• If budget weren't a limit, what would you build?
• If time weren't a factor, what would you do?
• If technology could do anything, what would be ideal?

Then work backwards: what fraction of that ideal solution is achievable within constraints?

Generate one option from constraint removal.

Technique 4 — Cross-domain transfer

How does [unrelated field] solve a structurally similar problem?

Domains to borrow from:

• Nature — evolution, ecosystems, ant colonies, immune systems
• Games — incentive design, progression systems, feedback loops
• Architecture — load-bearing structures, modularity, public vs. private space
• Music — composition, improvisation, harmony, rhythm
• Cooking — mise en place, flavour balancing, timing, presentation
• Transportation — routing, scheduling, hub-and-spoke, last-mile

Pick one domain. Identify the structural similarity. Extract the solution pattern.

Generate one option from cross-domain transfer.

Technique 5 — Worst idea first

What's the absolute worst solution you could propose?

Write it out. Then examine it seriously:

• What's accidentally interesting about it?
• What assumption does it violate that might be worth violating?
• What kernel of a good idea is hiding in the bad idea?

Generate one option by finding the good idea inside the bad one.

Quality bar for options

Every option must pass all three tests:

1. Distinct — it represents a fundamentally different approach, not a variation of another option
2. Feasible — it could actually work given real constraints (or with clearly identified constraint changes)
3. Specific — it's concrete enough to evaluate. "Use AI" is not an option. "Train a classifier on historical support tickets to auto-route incoming tickets to the right team, reducing triage time from 4 hours to 15 minutes" is an option.

If you have fewer than 5 options that pass all three tests, go back to the techniques and push harder. If you have more than 8, you're probably not being distinct enough — consolidate.

Step 4 — Evaluate honestly

For each option, provide:

### Option [N]: [Descriptive name]

**One-paragraph description:**
[What this option is and how it works]

**What's genuinely good about it:**
- [Specific strength — not "it's innovative" but "it reduces the feedback loop from days to minutes"]

**What's the biggest risk:**
- [The most likely way this fails]

**Effort to implement:**
[Low / Medium / High] — [one sentence justifying the estimate]

**Reversibility:**
[Easy to undo / Hard to undo / Irreversible] — this matters for decision-making

**Combinability:**
[Can elements of this option combine with other options? Which ones and how?]

Evaluation rules

• Be genuinely honest about pros and cons. Advocacy for a favourite kills the value of divergent thinking.
• "Innovative" is not a pro. "Reduces cost by 60%" is a pro.
• "Risky" is not a con. "Requires 3 months of engineering time with uncertain demand" is a con.
• The effort estimate must be realistic, not optimistic. Double your first instinct.

Step 5 — Combine and synthesise

After evaluating individually, look for combinations:

• Can the best elements of two options merge into something better?
• Can one option serve as the short-term solution while another is the long-term target?
• Can one option be the "safe bet" while another is the "high upside bet" — and can you pursue both?

If a combination emerges that's stronger than any individual option, describe it as a new hybrid option.

Step 6 — Output

Present the results in this order:

1. Problem reframing

Show all three reframes. State which was selected and why.

2. Options (the full set)

Each option with the evaluation from Step 4, numbered and named.

3. Recommended path

Which option (or combination) to pursue and why. Structure as:

**Recommendation:** [Option name or combination]

**Why this one:**
- [Primary reason tied to the problem and constraints]
- [Secondary reason]

**What to do first:**
- [Immediate next step — make it concrete]

**What to watch for:**
- [Early signal that this is working or not working]

4. Wild card

One option that is unconventional, possibly uncomfortable, but worth serious consideration. Explain why it shouldn't be dismissed.

**Wild card: [Name]**
Why it seems wrong: [the obvious objection]
Why it might be right: [the non-obvious argument]
When to revisit: [under what conditions this becomes the best option]

Rules

• The goal is options, not answers. Resist the urge to converge too early. Step 3 is about expanding the solution space; Step 5 is about converging.
• Genuinely different means genuinely different. Five variations on "build a web app" is one option, not five. Push past the obvious.
• Every option must be concrete enough that someone could say "yes, let's do that" and know what happens next.
• The wild card is mandatory. It's a hedge against groupthink and conventional wisdom. The best ideas often start as the weird one.
• Don't filter ideas by what's been tried before. If something was tried and failed, ask why it failed — the reason might no longer apply.
• Quantity enables quality. Generate broadly in Step 3, evaluate ruthlessly in Step 4. Mixing generation and evaluation kills creativity.
• Name every option. A named option is easier to discuss, reference, and decide on than "Option 3." Names should be descriptive and memorable.

Related Skills

• /council — when creative exploration produces multiple strong options, convene a council to debate which to pursue.
• /first-principles — when creative options feel constrained, decompose the problem to find assumptions that can be removed.
• /red-team — stress-test the recommended option before committing.