---
title: Sim Matrix
category: product
entity_type: agent
price: $79
canonical: https://forgehouse.ai/agents/sim-matrix/
lang: en
hreflang_alt: https://forgehouse.ai/tr/ajanlar/sim-matrix/
last_updated: 2026-06-20
---

# Sim Matrix

> Pre-production QA and quality gate

A pre-production quality gate that tests UI with simulated user personas before launch. It catches accessibility, Core Web Vitals and conversion breaks, and returns a clear PASS or FAIL. It tests the specific symptom, not just the status code: a fixed dropdown gets opened, an overflowing layout gets measured, across three breakpoints with real scrolling, because 'the page loads' has never meant 'the fix works'.

## Use cases
- Validating a page before it goes live
- Measuring Core Web Vitals and Lighthouse
- Auditing accessibility and keyboard navigation
- Catching layout breaks across breakpoints
- Verifying a specific bug fix actually worked

## Benefits
- Ship with confidence: breaks caught before users see them
- Real measurement, not 'looks fine on my screen'
- Accessibility and performance verified, not assumed
- A clear PASS/FAIL instead of vague opinions

## What’s included
- Pre-production gate with simulated personas: returns PASS or FAIL, never 'looks fine'
- Three-breakpoint render verification with real scroll, screenshot evidence
- Accessibility sweep: contrast math on real backgrounds, focus states, ARIA
- Specific-symptom testing: the fixed dropdown gets opened, not just loaded
- Console-error zero policy across every tested page
- Regression checks that previously-working features still work

## Who it’s for
Teams that want a hard quality gate before anything reaches production.

## How it runs
An LCP of 2.8 seconds is not close enough, it is a fail. This gate measures the deployed URL through four personas and hard thresholds, writes no code, and returns nothing but a verdict with the numbers attached.
1. Intercepts the handoff first: reads which files changed, which pages are affected and what the mission brief promised, then builds a simulation plan deciding which personas, pages and tests apply. It writes no code and fixes nothing; its only output is a verdict.
2. Targets the deployed preview or production URL, never a local server, and measures through browser tooling: a Lighthouse audit for Core Web Vitals, screenshots at 375, 768 and 1440 pixels, a console error sweep and a scan for failed network requests.
3. Simulates four personas against the page: a rushed mobile visitor with three seconds of patience, a desktop researcher, an elderly user on a tablet checking font size and keyboard navigation, and a search bot reading meta tags, schema and canonicals.
4. Issues the verdict in two stages. Stage one checks spec compliance against the mission brief's done criteria; a stage-one fail ends the run immediately. Stage two applies hard thresholds: LCP under 2.5 seconds, INP under 200 milliseconds, CLS under 0.1, contrast at least 4.5 to 1, touch targets at least 44 pixels, zero console errors and zero failed requests.
5. Never bends a threshold: an LCP of 2.8 seconds is not 'close enough', it is a fail. Every pass or fail line carries a real measurement; if a tool was unavailable, that metric is marked unverified rather than waved through.
6. On fail, returns the broken criteria with current value versus threshold and a suggested fix so the main thread can re-dispatch to the right builder, with a maximum of three fix-and-retest cycles before escalation. On pass, the change proceeds to final deployment verification.

## FAQ
### My site is already live, is sim-matrix only useful before launch?
It's designed as a pre-production gate, but it runs against any URL. Most teams use it before a deploy and after a bug fix, because it tests the specific symptom, the fixed dropdown gets opened, the overflowing layout gets measured. The verdict is still a hard PASS or FAIL.

### We already gate deploys on Lighthouse scores, what does persona testing catch that the numbers miss?
Lighthouse gives you scores; sim-matrix tests behavior. It renders three breakpoints with real scrolling and screenshot evidence, runs simulated personas through the page, computes contrast on real backgrounds, and enforces a zero-console-error policy plus regression checks on features that worked before. You get a verdict, not a number to interpret.

### Will sim-matrix fix the problems it finds?
No. It verifies and never writes code. A FAIL comes back with the specific break: which breakpoint, which element, what the screenshot shows, and the fix goes back to whoever built the page.

## Price
$79, one-time, no subscription. VAT included.

Related guide: [AI code review and developer workflow](https://forgehouse.ai/guides/ai-code-review/)