agency-debate

Great Minds Agency — Debate

Run a structured debate between Steve Jobs and Elon Musk with automatic Rick Rubin essence distillation.

Context

The debate is the first creative phase of the pipeline. Its purpose is to produce locked strategic decisions before any planning or building begins. Without debate tension, plans drift toward safe mediocrity.

Instructions

Step 0: Setup

1. Read the topic or PRD: $ARGUMENTS
2. Determine the project slug (lowercase, hyphens, no special chars)
3. Create the output directory: rounds/{slug}/

Step 1: Round 1 — Stake Positions

Launch both agents (ideally in parallel via Agent tool):

Steve Jobs (write to rounds/{slug}/round-1-steve.md)

Steve reads the PRD and stakes clear positions on ALL of these sections:

# Round 1 — Steve Jobs

## Design Philosophy
What aesthetic and interaction principles should govern this product?

## Architecture Opinion
Where the architecture must serve the experience (not the other way around).

## Naming & Brand
Product name, tagline, voice. What does it FEEL like?

## UX Flow
The critical user journey. What is the first 60 seconds?

## What to Cut
Features, screens, or concepts that dilute the vision. Be ruthless.

## What to Add
The one thing nobody asked for that makes it magical.

Each section must have a clear, defended position — not a list of options.

Elon Musk (write to rounds/{slug}/round-1-elon.md)

Elon reads the PRD and stakes clear positions on ALL of these sections:

# Round 1 — Elon Musk

## Design Philosophy
Where design choices create engineering debt or unlock scale.

## Architecture Opinion
Tech stack, data model, infrastructure. First-principles reasoning.

## Naming & Brand
Does the name scale? Is it defensible? Global viability.

## UX Flow
Friction analysis. Where do users drop off? What needs to be instant?

## What to Cut
Scope that delays launch. What is NOT in v1?

## What to Add
The force multiplier — the feature that makes growth inevitable.

Each section must have reasoning, not just assertions.

Step 2: Round 2 — Challenge and Defend

Each agent reads the OTHER's Round 1 and writes a challenge response.

Steve reads Elon's R1 (write to rounds/{slug}/round-2-steve.md)

# Round 2 — Steve Jobs (Response to Elon)

For EACH of Elon's 6 sections:
## Re: {Section Name}
**Elon said**: {one-sentence summary}
**My challenge**: {specific disagreement with reasoning}
**My position holds because**: {defense}
**Concession (if any)**: {what Steve acknowledges Elon got right}

Elon reads Steve's R1 (write to rounds/{slug}/round-2-elon.md)

# Round 2 — Elon Musk (Response to Steve)

For EACH of Steve's 6 sections:
## Re: {Section Name}
**Steve said**: {one-sentence summary}
**My challenge**: {specific disagreement with reasoning}
**My position holds because**: {defense}
**Concession (if any)**: {what Elon acknowledges Steve got right}

Step 3: Decisions Document

Read all 4 round files. Write rounds/{slug}/decisions.md:

# Locked Decisions — {Project Name}

For each of the 6 areas:
## {Area}
**Decision**: {the locked choice}
**Proposed by**: Steve / Elon / Synthesis
**Reasoning**: {why this won}
**Dissent**: {what the other side still disagrees on}

## MVP Feature Set
{Bulleted list of what ships in v1}

## Deferred to v1.1
{What was cut and why}

## File Structure
{Proposed directory/file layout for the build phase}

Step 4: Rick Rubin — Essence Check (Auto-Trigger)

After decisions.md is written, spawn a haiku sub-agent as Rick Rubin:

Agent(model: "haiku", subagent_type: "rick-rubin-creative",
  prompt: "Read rounds/{slug}/round-2-steve.md and rounds/{slug}/round-2-elon.md
  and rounds/{slug}/decisions.md.
  Strip everything to essence. Answer ONE question:
  'What are the 3 things that actually matter?'
  No filler, no diplomacy, no corporate language.
  For each of the 3 things, write ONE sentence explaining why it matters.
  Write to rounds/{slug}/rick-rubin-essence.md")

Rick's output feeds into /agency-plan — the plan must address his 3 essentials or explicitly justify why not.

Step 5: Report to User

Summarize:

• The 3-5 most important locked decisions
• Key tensions that remain
• Rick Rubin's 3 essentials
• Tell user to run /agency-plan to proceed

Output Files

File	Content
rounds/{slug}/round-1-steve.md	Steve's 6-section position paper
rounds/{slug}/round-1-elon.md	Elon's 6-section position paper
rounds/{slug}/round-2-steve.md	Steve challenges Elon's positions
rounds/{slug}/round-2-elon.md	Elon challenges Steve's positions
rounds/{slug}/decisions.md	Locked decisions with reasoning
rounds/{slug}/rick-rubin-essence.md	The 3 things that actually matter

Philosophy

The debate produces positions under pressure — not consensus, not compromise. The best ideas survive challenge. The goal is clarity, not agreement.

Anti-Patterns

• Do NOT let agents agree on everything — force tension
• Do NOT skip sections — all 6 areas must be addressed in both rounds
• Do NOT write vague positions like "we should consider..." — stake a claim
• Do NOT skip Rick Rubin — the essence check catches bloat before it reaches planning