Skip to content
Research Hub
Research · Visual Management

Visual IntelligenceSystem

Agentic visual asset management for AI-native creators. Scan, audit, and fix every image across your codebase with a single command.

408

Images Indexed

268

Pages Mapped

76/100

Health Score

6

Architecture Layers

The Visual Chaos Problem

Modern AI-native projects accumulate visual assets faster than any manual process can track. FrankX.ai's initial audit revealed the scope of this challenge.

333

orphaned images

Images present in the repository with unknown usage status. Every orphan is wasted storage and potential confusion for contributors.

2

placeholder SVGs on flagship posts

Default placeholder graphics displayed on high-traffic blog posts instead of proper featured images, reducing visual authority.

0

systematic auditing before VIS

Every visual decision was manual. There was no automated pipeline to detect issues, enforce brand consistency, or measure visual health.

6-Layer Architecture

Each layer handles a distinct responsibility, from raw file discovery through automated remediation. Layers compose into a single CLI command or integrate individually via API.

01

Scanner

Recursively indexes every image across the codebase. Maps file paths, dimensions, formats, and byte sizes into a structured inventory.

408 images indexed
02

Auditor

Cross-references images against page usage, alt text coverage, format optimization, and responsive sizing rules.

268 pages mapped
03

Brand DNA

Validates visual consistency against brand palette, aspect ratios, naming conventions, and quality thresholds.

Per-image grading
04

Intelligence

Scores overall visual health, detects orphans, duplicates, oversized assets, and missing responsive variants.

Health score engine
05

Integration

Connects with ACOS, Claude Code, Cloudinary, and n8n to automate fixes, uploads, and CDN optimization.

6 integrations
06

CLI

Single-command interface: scan, audit, fix, report. Generates markdown reports and JSON inventories for CI pipelines.

vis scan --fix

Case Study: FrankX.ai

The Visual Intelligence System was built and validated against the FrankX.ai production codebase. Here are the measured results from the first full audit cycle.

Health Score

Before

1/100

After

76/100

Placeholders Replaced

Before

2 active

After

0 remaining

Duplicates Fixed

Before

3 detected

After

0 remaining

Images Indexed

Before

0

After

408

Pages Mapped

Before

0

After

268

Orphaned Images

Before

333 unknown

After

333 catalogued

How It Works

Three steps from visual chaos to measured health. Each step runs independently or chains into a single pipeline.

01

Scan

Point VIS at any directory. It recursively discovers every image file, extracts metadata, and builds a structured inventory in seconds.

02

Audit

Cross-reference images against every page. Detect orphans, missing alt text, oversized files, placeholder SVGs, and brand inconsistencies.

03

Fix

Auto-replace placeholders, compress oversized assets, generate responsive variants, and update references across the codebase.

Integrations

VIS connects with the tools already in your stack. Each integration extends the pipeline with specialized capabilities.

Claude Code

AI-powered codebase operations

nanobanana

AI image generation pipeline

Cloudinary

CDN optimization and transforms

n8n

Workflow automation triggers

Slack

Audit report notifications

Get Started with VIS

The Visual Intelligence System ships as part of ACOS. Install it, run a scan, and see your visual health score in minutes.