D
design.x1000.aiAI Design OS

Documentation

Operating Method

How m558 works, how to integrate it into your workflow, and how it connects to the LIGHT HOPE ecosystem standards.

What is m558?

m558 is a design coordination layer for the LIGHT HOPE / X1000 ecosystem. It turns visual design rules into structured markdown that coding agents can read and execute consistently.

The problem it solves: AI agents generate UI quickly, but produce inconsistent output across sessions. Without a shared design reference, every page looks slightly different. m558 eliminates that drift.

# The core model

DESIGN.md → brand rules (what is allowed)

persona/*.md → emotional tone (how it should feel)

prompts/*.md → execution spec (what to produce)

= Coherent UI every time

The Four Layers

L1 — Design System

design-systems/{domain}/DESIGN.md

Defines the brand rules for a product domain. Colors, typography, spacing, component patterns, what to avoid. This is the law.

L2 — Design Persona

personas/{persona}.md

Defines the emotional register. Not what colors to use — how they should feel. The difference between "uses blue" and "uses blue like Apple would".

L3 — Execution Prompt

prompts/{tool}-v{n}.md

The operational instruction for a specific coding agent. Tells it what files to read, what to produce, and how to validate success.

L4 — Page Brief

demos/{name}/brief.md

A specific page generation spec. Describes the page, audience, sections, constraints, and expected CTA. Plug it into the execution prompt.

How to use in a project

Option A — Direct file reference

# In Claude Code session:
Read these files before writing any UI:
1. /path/to/m558/design-systems/lighthope/DESIGN.md
2. /path/to/m558/personas/airbnb-warm.md

Then build: {your page}

Option B — Copy to project root

cp m558/design-systems/lighthope/DESIGN.md your-project/
cp m558/personas/airbnb-warm.md your-project/
# Then reference relative paths in prompts

Option C — MCP Memory Bridge (V5.4) ✅ Live

# m558 is registered in m556 MCP (project #23)
# Agents can fetch m558 context via:
get_project_overview("m558")
get_project_structure("m558")
get_project_issues("m558")

⚠️ Existing Project Integration (v1.1.0)

# For EXISTING projects with established brands:
# 1. Read target project's tailwind.config + globals.css FIRST
# 2. Use DESIGN.md for layout patterns, NOT to override brand colors
# 3. Persona = emotional tone within existing brand, not new palette
# See "Integration Guide" in each DESIGN.md for details

V5.3 + V5.4 Compliance

m558 follows the LIGHT HOPE Ecosystem Standard V5.3 (Stability) and V5.4 (AI OS Integration).

Docker Compose

✅ V5.3 template

Memory limit

✅ 1GB cap

Restart policy

✅ unless-stopped

/api/health

✅ JSON response

Log rotation

✅ 10MB × 3

Port convention

✅ m558 → 8558

MCP snapshot

✅ m556 registered

Visual audit

✅ Playwright-based

Skills registry

⏳ V2 planned

README > 100 lines

✅ 248 lines

What's coming

V1.0.0Shipped
7-page static site4 personas3 DESIGN.md systems3 execution prompts3 demo briefsDocker + Cloudflare
V1.1.0Current
Integration Guide (new vs existing)Prompt STEP 0 checkStandalone deploy fixm556 MCP livePlaywright visual auditFirst production practice (m510)
V2.0.0Backlog
Interactive playgroundLive persona switchingURL → DESIGN.md extractionDesign consistency scorer
V2.1.0Backlog
m495 skills integrationdesign-md-generator skillpersona-switcher skillbrand-audit skill

Read one DESIGN.md. Ship one page.

The fastest way to grok m558 is to run one execution prompt on a real page. Try the playground to see four personas on the same content, then use the Claude Code prompt.