A ROMULII™ framework · a no-think, just-build tool
Everyone's prompt builds one.
This framework builds 30.
Drop the bundle into Claude, describe your app in one sentence, type /rrStart. About 5 prompts and 1–2 hours later you have 30 clickable, phone-framed prototypes to compare, pick from, and build on.
▶ Or click through a live 30-build demo- Tested end-to-end on Claude
- Works on Claude Pro
- Browser-first, terminal optional
- One-time payment
The problem
One prompt. One wireframe. Now what?
A single prompt gives you a single interpretation. You take the first thing that looks okay — and never see the directions you didn't ask for.
You explore too few directions before committing.
You settle on the first acceptable idea.
Important screens and flows get overlooked.
You burn time re-prompting to nudge one output.
There's no systematic product-design process underneath.
You end with isolated screens, not a product vision.
The big idea
Stop prompting. Start exploring.
The value isn't one wireframe. It's exploring the product space around the idea — then choosing the strongest direction instead of settling for the first. Clients see the result, not the process.
See the difference
One idea. 30 ways to see it.
30 visual directions: 21 approved UI style boards plus 9 derived from your own spec — each with light and dark. All 30 sit inside phone frames on one page you double-click. No server, no keys, no setup.
Try it live
Don't take our word for it. Click through all 30.
Real step 3 output for one idea — Tat's Cake Studio, a cake-ordering app. 30 directions, two screens each: the home and the pickup calendar. Use the same controls your build ships with.
- All → Home
- All → Pickup calendar
- Shortlisted only
- Copy picks
- Check fit
- GPT imagery on / off
How it works
Functional specs. Design specs. Build 30. Select.
A locked pipeline of prompts, rulebooks and templates. Nothing to read first: attach the bundle, add your app name and description, type /rrStart — the AI configures itself and starts you.
Optional: expand your idea
Use any free AI — free ChatGPT works — to turn your messy sentence into a fuller brief. Free and not counted as a prompt.
optional · free · uncountedFunctional specs — the foundation
Your idea, tagline and name go in. One archive comes out: 23 documents including the product bible, the PRD and the full functional spec.
/rrStart · one archive outDesign specs — 30 directions
30 complete visual directions (21 approved UI style boards + 9 derived from your spec), each distinct in content and layout, each with light and dark.
design system · tokens · flowsBuild 30 — the prototypes
30 prototypes, two screens each: the home screen and the screen hardest to get right. Every one hand-built inside a phone frame, all in one page.
no server · no keys · no setupComplete, audit, select, expand
The build is filled in and checked against the spec. Pick your direction, then run step 4 to explore more sections and polish it up. Going live unlocks here in editions that include it.
checked against the specThen it's your call.
Scrap it (4–5 prompts, not 5 weeks) · continue it yourself · hand it to a dev with HANDOVER.md · try a new idea.
What a build gives you
Compare 30. Shortlist. Expand the winner.
The quick MVP build is for choosing. The polish comes after you've chosen.
Two screens per direction
Each build makes the home screen plus the screen hardest to get right — enough to judge a direction without paying for 30 full apps.
A comparison page built in
Flip every build to the same screen, keep a shortlist, copy your picks and check fit.
Expand after you select
Run step 4 on your selection(s) to explore more sections and polish up.
Add real imagery
Polish a build in GPT, or use the Higgsfield MCP in Claude to add images to the wireframes.
What's in the box
A framework, not an app.
You run the prompts; the AI builds your product. Here's what ships, and which edition includes it.
The locked prompt pipeline A · B · C
Identical steps, manifests and checksums in every edition.
30 clickable prototypes per idea A · B · C
All in phone frames, in one page you double-click.
Functional specification files step 1 output
Product bible, PRD, functional spec, data model, user journeys, MVP scope, security model and more.
Design specification files step 2 output
Design system, tokens schema, component primitives, motion, navigation patterns, accessibility, responsive guidelines.
Romulii OS framework B · C
Rulebooks, gates, stages and the go-live path.
One-click terminal Launcher C
Stamps a fresh, correctly-wired project folder per app — Windows and Mac launchers — so repeat builds skip setup.
START-HERE, user guide, HANDOVER.md
Fully self-serve. HANDOVER.md tells a developer exactly what exists and what to wire.
LICENSE.md
Commercial use for your apps. Your download is the complete product, as it exists at purchase.
Step 1 output — functional specification files24 files · one archive
Step 2 output — design specification filesconcepts folder + 26 files
Full edition (C) bundle — browser, terminal + bonusproject folder · pipeline files · launcher
Project folder
.claude/ADAPTERS/assets/blueprints/data/docs/memory/prompts/reports/research/src/ .env.example.gitignore.mcp.jsonCLAUDE.mdCOMPLETION-PROTOCOL.mdCOST-CONTROL.mdCOST-REPORT.mdDECISIONS.mdDESIGN-SYSTEM.mdEXTENSIBILITY.mdFOLDER-MAP.mdforge.config.jsonforge-status.jsIMPROVEMENTS.mdLAUNCH (MAC).commandLAUNCH (WINDOWS).batLEGAL-PACK.mdLICENSE.mdPROJECT_STATE.mdREACH-CONTROL.mdREADME.mdRENAME (MAC).commandRENAME (WINDOWS).batRename-App.ps1rename-app.shRISK-CONTROL.mdROMULII-SYSTEM-PROMPT.mdRRSTART.mdSETUP-MCP-AND-SKILLS.mdSHIP-CHECK.mdSTACK-UPGRADES.mdSTART-HERE.mdSTATUS (MAC).commandSTATUS (WINDOWS).batPipeline files
ADAPTERS/CLAUDE.mdCOMPLETION-PROTOCOL.mdCOST-CONTROL.mdCOST-REPORT.mdDECISIONS.mdIMPROVEMENTS.mdLEGAL-PACK.mdLICENSE.mdmodule-05 … module-47.md (×20)PIPELINE-CHECKSUMS.txtPIPELINE-LOCK.mdPROJECT_STATE.mdREACH-CONTROL.mdREADME.mdRESEARCH-CHAT-SETUP.mdRISK-CONTROL.mdROMULII-SYSTEM-PROMPT.mdRRSTART.mdSHIP-CHECK.mdLauncher
_template/Create-RomuliiApp.ps1create-romulii-app.shNEW-ROMULII-APP (MAC).commandNEW-ROMULII-APP (WINDOWS).batREADME-LAUNCHER.mdWho it's for
For people who need more than one good idea.
Give clients variety, or compare and select the strongest direction yourself. Skip blank-canvas wireframes and start from a clear UI direction.
Built for beginner, intermediate and advanced users · bundles for chat, terminal and code users · dyslexia & ADHD-first, by a dyslexic author.
“I built this for myself first — to quickly turn ideas into something useful for clients, friends and family. Then I realized it could help a lot more people.”
Remus Romulus, ROMULII™
Read before buying
Before you buy.
Everything you need to know, up front. Sold like a book: a finished system, not a subscription. Your download is the complete product and will not be updated.
What you need
- A Claude subscription (claude.ai). At the time of writing, Pro is enough — every prompt runs on the standard tiers.
- Model: at the time of writing, Claude Opus 4.8 is the recommended pick for runtime, token use and reliability. Explore on it; polish with a more advanced model after step 3. Later Claude releases may change this pick.
- Max plan (optional): use the highest creative tier for the final polish. Never a requirement.
- Optional: free ChatGPT to expand your idea before step 1.
- A framework, not an app: you run the prompts; the AI builds your product. AI output varies run to run; no specific result is guaranteed.
To go live B · C
- Free-tier Vercel account.
- Free-tier Supabase account.
- A small Anthropic API budget, which the OS is designed to cap.
- Going live applies only where your edition includes it. Chats-only (A) is browser-only.
- Third-party plans, prices and terms are set by those providers and may change; any costs with them are yours.
Platform
- Built and tested on Windows.
- macOS: untested, but provisions were made (Mac .command launchers ship alongside Windows .bat).
- Browser-first. The core pipeline runs in the Claude browser — no terminal needed.
Support & updates
- Included: the full system in your edition, as it exists at the time of purchase.
- Not included: upgrades, updates, fixes, future versions or ongoing support, beyond any required by law. None are promised or scheduled. Any new version, if ever released, is a separate purchase; no free or discounted upgrade is included or promised.
- No consulting or support. Fully self-serve — START-HERE, the user guide and every prompt explain themselves.
License & refunds
- One named individual per purchase. You may deliver builds for your own clients — freelance and consulting use included.
- Not for teams. Not shared or pooled across people, not for a team or organization under one purchase. Each additional person needs their own purchase.
- Your apps are 100% yours, commercial use included. The OS itself may not be resold or redistributed.
- Built apps ship with a small removable “© Remus Romulus / ROMULII™” footer (one config line removes it) and non-linking framework identifiers, disclosed in LICENSE.md.
- Refunds: digital download, delivered immediately. At checkout you consent to immediate delivery and acknowledge that under EU/UK digital-content rules the right of withdrawal ends once the download begins.
- Checkout and payment are handled by the store provider (Lemon Squeezy) as merchant of record; their refund policy governs the transaction — see it at checkout. By purchasing you accept the LICENSE.md included in the bundle.
- Nothing here affects any rights you have under applicable consumer law.
Claude is a product of Anthropic; ChatGPT, Vercel and Supabase belong to their respective owners. Wireframe Killer™ is independent and not affiliated with or endorsed by any of them.
Honest note: AI output varies
Every AI — any brand, model, version or effort setting — hallucinates and varies run to run. Romulii OS constrains this hard (locked steps, machine-readable manifests, checksums) and enforces quality floors, but no run is guaranteed identical or error-free. Results depend on the model, plan and your inputs. You review what it builds; the framework makes that review fast. Expect to review and iterate — that's by design, not a defect.
Three editions
Pick your edition.
Same locked pipeline in every edition — identical steps, manifests and checksums. Editions differ by tooling, never by build quality.
30-Wireframe Killer (Browser files only)
For designers. The locked pipeline in the browser — the 30 prototypes.
- Locked prompt pipeline
- 30 clickable prototypes per idea
- Functional + design spec outputs
- Browser only — no setup
- Romulii OS · rulebooks, gates, stages
- Go-live path
- Terminal launcher
30-Wireframe Killer (Browser + OS framework)
Adds the full operating system — rulebooks, gates, stages, go-live.
- Everything in Chats-only
- Full Romulii OS framework
- Rulebooks, gates & stages
- Go-live path (Vercel + Supabase)
- Terminal launcher
30-Wireframe Killer Full (All + Launcher)
For repeat builders. Adds the one-click terminal launcher.
- Everything in Chats + OS
- One-click terminal launcher
- Fresh, wired project folder per app
- Windows + Mac launchers
- Browser, terminal + bonus
| Feature | A · Chats-only | B · Chats + OS | C · Full |
|---|---|---|---|
| Locked pipeline, manifests, checksums | ✓ | ✓ | ✓ |
| 30 clickable prototypes per idea | ✓ | ✓ | ✓ |
| Runs in the Claude browser | ✓ | ✓ | ✓ |
| Romulii OS — rulebooks, gates, stages | – | ✓ | ✓ |
| Go-live path | – | ✓ | ✓ |
| One-click terminal launcher | – | – | ✓ |
| Build quality | Same | Same | Same |
License: one named individual per purchase. Client work is fine; teams need one purchase per person. Read before buying →
Why Wireframe Killer™
Because the first idea isn't always the best idea.
Explore breadth before you commit depth. With 30 directions on the table you're choosing the strongest, not defending the first.
Exploration
Compare 30 directions before building.
Validation
Test ideas before development.
Creativity
Discover concepts you wouldn't design alone.
Decision making
Choose the strongest direction faster.
Flexibility
Build, scrap, continue, or hand off.
Agency advantage
Handle more projects with the same workflow. The client sees one; you explored 30.
FAQ
Questions, answered.
The product
What exactly is Wireframe Killer™?
A framework, not an app. Romulii OS is an operating system of prompts, rulebooks and templates that takes one messy sentence and turns it into a field of clickable prototypes. You run the prompts; the AI builds your product.
How long does it take, and how many prompts?
Idea to 30 clickable prototypes in about 1–2 hours (model dependent), in about 5 prompts.
What does each of the 30 prototypes include?
Two screens: the home screen and the screen hardest to get right, inside a phone frame. All 30 live on one page you double-click — no server, no keys, no setup. After you select, step 4 expands more sections.
Do I need design or coding experience?
No. It's built for beginner, intermediate and advanced users. Attach the bundle, add your app name and description, type /rrStart — nothing to read first.
Can I add real images to the wireframes?
Yes — polish a build in GPT, or use the Higgsfield MCP in Claude to add images.
What happens after the 30 are built?
Your choice: scrap it, continue it yourself (step 4 walks a beginner through picking a design, finishing it and going live), or hand it to a developer — HANDOVER.md tells them what exists and what to wire.
Requirements
Which Claude plan and model do I need?
At the time of writing, Claude Pro is enough. Opus 4.8 is the recommended model at the time of writing; Max users can optionally use the highest creative tier for final polish. Later Claude releases may change this pick.
Does it work on Mac?
It was built and tested on Windows. macOS is untested, but provisions were made, including Mac launchers. The core pipeline runs in the Claude browser.
What do I need to go live?
In editions that include it (B and C): free-tier Vercel and Supabase accounts, plus a small Anthropic API budget that the OS is designed to cap. Third-party plans, prices and terms may change; any costs with them are yours.
Will two runs give the same result?
No. AI output varies run to run. The pipeline enforces quality floors and checks, but no run is guaranteed identical or error-free. Review and iterate — that's by design.
Buying
What's the difference between the editions?
Chats-only (A): the locked pipeline, browser only, the 30 prototypes. Chats + OS (B): adds rulebooks, gates, stages and go-live. Full (C): adds the one-click terminal launcher for repeat builders. Same pipeline and build quality in all three. Compare editions.
Is it a subscription?
No. One-time payment. It's sold like a book: a finished system, as it exists at purchase. Your download is the complete product and will not be updated — no upgrades, updates, fixes or support, beyond any required by law. Any new version, if ever released, is a separate purchase.
Can my team share one purchase?
No. One named individual per purchase. Not shared or pooled, and not for a team or organization under one purchase — each additional person needs their own.
Can I use it for client and commercial work?
Yes. Your apps are 100% yours, commercial use included, and you may deliver builds for your own clients. The OS itself may not be resold or redistributed. Built apps carry a small removable footer (one config line) and non-linking framework identifiers, disclosed in LICENSE.md.
Is there support?
The OS is fully self-serve: START-HERE, the user guide and every prompt explain themselves. No consulting or support is included.
What's the refund policy?
This is a digital download delivered immediately. At checkout you consent to immediate delivery and acknowledge that, under EU/UK digital-content rules, the right of withdrawal ends once the download begins. Lemon Squeezy is the merchant of record, and their refund policy governs the transaction — see it at checkout.
How do I get my files, and where is payment handled?
Checkout runs through Lemon Squeezy (Stripe), and the download is delivered immediately. Your download is the complete product and will not be updated. By purchasing you accept the LICENSE.md in the bundle. Nothing here affects any rights you have under applicable consumer law.
Is Wireframe Killer™ affiliated with Anthropic or Claude?
No. Claude is a product of Anthropic; ChatGPT, Vercel and Supabase belong to their respective owners. Wireframe Killer™ is independent and not affiliated with or endorsed by any of them.
Your next idea deserves more than one wireframe.
Everyone's prompt builds one. Wireframe Killer™ builds 30 — then you choose, refine and build.
Choose your editionOne-time payment · Lemon Squeezy · Unlimited builds