THE BLUEPRINT PATH
You already have a complete PRD. Drop it in and let the crew execute.
You arrive with orders already written. Good. I'll validate your frontmatter, scan for conflicts, and prepare the crew. You don't need the wizard's hand-holding — you need the bridge crew's precision.
PicardWHAT YOU'LL NEED
A finished PRD at docs/PRD.md — with valid YAML frontmatter. The only required field is name. Optional but recognized: type, framework, database, deploy, auth, payments, workers. Need help writing one? Use /prd to generate it.
VoidForge — create a project (requires Node 20+):
npx thevoidforge init my-projectClaude Code — Anthropic's coding CLI. Install with: curl -fsSL https://claude.ai/install.sh | bash
Optional: Supporting documents — ADRs in docs/adrs/, project directives in docs/PROJECT-DIRECTIVES.md, operations playbook in docs/OPERATIONS.md, reference materials in docs/reference/.
DROP YOUR SPEC
Copy your PRD into the project:
cp ~/my-spec.md docs/PRD.mdIf you have supporting documents, place them where the agents expect them:
docs/PROJECT-DIRECTIVES.md— project-specific coding rules (appended to CLAUDE.md)docs/adrs/*.md— architecture decisions Picard referencesdocs/OPERATIONS.md— operational constraints Sisko referencesdocs/reference/*— any reference material available to all agents
I scan every corner of docs/ for supporting material. ADRs, directives, operations playbooks, reference files — anything that helps the crew build smarter. Drop it in docs/ and I'll find it.
WongRUN /BLUEPRINT
First, launch Claude Code in your project directory:
cd my-projectclaudeThis launches Claude Code in your project directory. Claude automatically reads the CLAUDE.md methodology file, which loads VoidForge's agents and commands. At the Claude Code prompt, type:
/blueprintHere's what happens under the hood (see /commands/blueprint for full detail):
- Picard validates — parses YAML frontmatter, checks required fields, extracts architecture
- Troi checks structure — does the PRD have features? data models? deployment section?
- Wong discovers docs — scans for supporting materials, loads them into context
- Directives merge — project-specific rules appended to CLAUDE.md (idempotent)
- Conflict scan — checks for structural contradictions (auth without database, workers on static hosting, etc.)
- Kusanagi provisions — sets up infrastructure from frontmatter (framework, database, deploy target)
THE CHALLENGE (OPTIONAL)
For extra confidence, add --challenge (inside Claude Code):
/blueprint --challengeBoromir reads your entire PRD and argues against it — expensive features, fragile integrations, schema gaps, deploy target mismatches. A 30-second argument saves a 3-hour refactor.
You can accept challenges (edit PRD) or override (proceed as-is).
One does not simply ship a PRD without questioning it. I find the expensive features you'll regret, the integrations that will break at 3am, and the schema decisions that paint you into a corner. Better to argue now than debug later.
BoromirBUILD AND DEPLOY
Once validated, run the campaign (inside Claude Code):
/campaign # Autonomous build, full roster (default)/campaign --interactive # Pause between missions for inspectionSisko reads the validated PRD, breaks it into missions, and executes end to end. Each mission: build, review, fix, commit.
When complete, Kusanagi deploys.
Picard validated your spec. Wong loaded your docs. Boromir argued and lost. Now I provision the target — Vercel, AWS, Railway, Docker, bare metal. Your blueprint becomes infrastructure.
KusanagiWHAT'S NEXT
Your spec is validated, built, and deployed. Now stress-test it before real users arrive. Run the Gauntlet — 30+ agents review every domain before you ship.
Once you're live, the growth engine handles SEO, ads, social, and outreach — so your launch doesn't end on day one.
For full command details, see the /blueprint command reference.
Don't have a spec yet? Start with the Wizard — it builds your PRD from scratch through an interactive conversation.