architect

Nexus Architect

System architecture toolkit for Claude Code and Codex. Claude Code uses this repository as two plugins with 52 skills; Codex uses the same skill files through AGENTS.md compatibility rules.

• architect (41 skills) — Legacy refactoring, greenfield design, database migration, consulting deliverables
• scalardb (11 skills) — ScalarDB application development toolkit

Installation

As a Claude Code Plugin (Recommended)

# 1. Add the marketplace
claude plugin marketplace add wfukatsu/nexus-architect

# 2. Install both plugins
claude plugin install architect@nexus-architect --scope user
claude plugin install scalardb@nexus-architect --scope user

After installation, commands are available as /architect:skill-name and /scalardb:skill-name.

To update to the latest version:

claude plugin update architect@nexus-architect
claude plugin update scalardb@nexus-architect

Manual Installation

# 1. Clone the repository
git clone https://github.com/wfukatsu/nexus-architect.git

# 2. Add as a local marketplace
claude plugin marketplace add ./nexus-architect

# 3. Install both plugins
claude plugin install architect@nexus-architect --scope user
claude plugin install scalardb@nexus-architect --scope user

Verify Installation

In a Claude Code session, type any command to confirm:

/architect:start
/scalardb:model

If the skills are recognized, the installation is successful.

Using with Codex

Codex can use the same skill files without installing Claude Code plugins.

# 1. Clone the repository
git clone https://github.com/wfukatsu/nexus-architect.git
cd nexus-architect

# 2. Optional Python dependencies
pip install -r requirements.txt

Open Codex at the repository root. AGENTS.md tells Codex how to translate Claude Code conventions:

• /architect:<name> -> skills/<name>/SKILL.md
• /scalardb:<name> -> skills/<name>/SKILL.md
• CLAUDE_PLUGIN_ROOT -> the repository root
• .claude/docs/* -> skills/common/references/*
• .claude/rules/* -> rules/*
• ${CLAUDE_PLUGIN_ROOT}/subagents/* -> skills/common/subagents/*

Then invoke the same command text in chat:

/architect:start ./path/to/target
/architect:pipeline ./path/to/target
/scalardb:model
/scalardb:review-code ./path/to/app

When a skill asks to use Claude tools, Codex follows these mappings:

Claude Code reference	Codex behavior
Read, Glob, Grep, LS	Use shell reads, rg, rg --files, find, or ls
Write, Edit, MultiEdit	Edit files with apply_patch
Bash	Run shell commands
AskUserQuestion	Present numbered choices in chat and wait for the reply
Task, Subagent	Run in the main Codex thread unless the user explicitly asks for sub-agents
WebFetch, WebSearch	Use Codex web access, Context7, or approved curl

After editing generated reports or Mermaid diagrams in Codex, run the hooks manually when relevant:

hooks/validate-frontmatter.sh reports/before/example/technology-stack.md
hooks/validate-mermaid.sh reports/before/example/codebase-structure.md

Claude Code continues to use the plugin metadata and slash commands unchanged. See Using Nexus Architect with Codex for the full Codex guide.

Quick Start

# Interactive workflow (recommended)
/architect:start ./path/to/target

# Automated full pipeline
/architect:pipeline ./path/to/target

# Individual skills
/architect:investigate ./path/to/target
/architect:analyze ./path/to/target
/architect:evaluate-mmi ./path/to/target

# ScalarDB development
/scalardb:scaffold
/scalardb:model
/scalardb:build-app

Commands

Orchestration

Command	Description
/architect:start	Interactively start system analysis and design
/architect:pipeline	Automated pipeline execution (supports --resume, --skip)
/architect:init-output	Initialize output directories

Requirements

Command	Description
/architect:define-requirements	Requirements definition: FR/NFR classification, data/transaction requirements, ScalarDB applicability (greenfield entry point)

Investigation & Analysis

Command	Description
/architect:investigate	Tech stack, structure, debt, DDD readiness survey
/architect:investigate-security	OWASP Top 10, access control assessment
/architect:analyze	Ubiquitous language, actors, domain mapping
/architect:analyze-data-model	Data model, DB design, ER diagrams

Evaluation

Command	Description
/architect:evaluate-mmi	MMI 4-axis qualitative evaluation
/architect:evaluate-ddd	DDD 12-criteria 3-layer evaluation
/architect:integrate-evaluations	MMI+DDD integration, improvement plan

Design