/project-launch — Parallel Founding Team

Launch all founding-team agents in parallel on your new project. What would take 40+ minutes sequentially finishes in under 10 minutes as one parallel burst.

Usage

/project-launch <description>
/project-launch "SaaS dashboard with Stripe billing, user auth, real-time data"
/project-launch --full    # adds performance-optimizer and api-guardian to the team

What Gets Launched in Parallel

Five specialist agents fire simultaneously, each with the same project description:

| Agent | Deliverable | |---|---| | architect | System design — components, data flow, boundaries, trade-offs | | pair-programmer | Approach validation — risks, wrong turns, questions to answer before coding | | security-reviewer | Threat model — auth, data exposure, OWASP risks for THIS project type | | tdd-guide | Test strategy — what to test first, test pyramid, critical paths | | ux-reviewer | UX blueprint — user flows, accessibility requirements, interaction risks |

With --full, also adds:

• performance-optimizer — scaling bottlenecks, caching strategy, query patterns
• api-guardian — API contract design, versioning strategy, breaking change risks

How It Works

• Read the project description from $ARGUMENTS
• If no description provided, ask for one in one sentence
• Spawn all 5 (or 7) agents simultaneously with run_in_background: true
• Each agent receives: the project description + its specialist brief
• Wait for all to complete
• Aggregate results into a structured Project Launch Report:

## PROJECT LAUNCH REPORT
Architecture (architect)
[findings]
Approach Risks (pair-programmer)
[findings]
Threat Model (security-reviewer)
[findings]
Test Strategy (tdd-guide)
[findings]
UX Blueprint (ux-reviewer)
[findings]

• After report: ask "Which area do you want to go deeper on first?"

Token Efficiency

Each agent gets a narrow, focused context — just the project description + its specialist instructions. No agent carries the full conversation history. The system prompt (rules + agents) is cached across all parallel agents. Result: 5× the analysis at roughly 2× the token cost of a single sequential session.

When to Use

• Starting any non-trivial project (3+ components, real users, production intent)
• Before writing a single line of code
• When you want to catch architecture mistakes before they're baked in

Arguments

$ARGUMENTS

Execute Now

Invoking this command is confirmation to proceed. Do not show this documentation to the user — execute immediately:

• Read $ARGUMENTS for the project description. If empty, ask: "What are you building? Describe it in 2–3 sentences."
• Once you have a project description, spawn all agents simultaneously using run_in_background: true:

• Wait for all agents to complete.
• Aggregate into the Project Launch Report format above.
• End with: "Which area do you want to go deeper on first?"