Reflexive Claude Code

Reflexive Claude Code

[English](README.md) | [繁體中文](README.zh-TW.md)

A Claude Code plugin marketplace for **skills-driven Agentic Context Engineering (ACE)** — build, analyze, and maintain agent systems with structured workflows.

What This Does

Reflexive Claude Code gives Claude Code a complete toolkit for managing its own agent system: CLAUDE.md, rules, skills, subagents, and hooks. Every component follows a structured creation process with automated quality review.

**Key capabilities:**

• Build agent systems from scratch — detect project maturity, explore workflows, plan architecture, then create all components in dependency order
• Analyze existing systems — 11-category weakness checklist covering routing, context management, security, rules health, and cross-tool migration
• Validate on every edit — PostToolUse hook checks frontmatter, broken links, orphaned files, and invalid variables in real time
• Enforce quality gates — 5 specialized reviewer agents (skill, CLAUDE.md, rule, hook, subagent) run after every creation or modification
• Advanced security architecture — AI-specific security patterns, eight-layer defense system, 30-second performance constraints, and intelligent reflection-driven learning

Installation

/plugin marketplace add wayne930242/Reflexive-Claude-Code
/plugin install rcc@rcc          # core ACE workflow
/plugin install aref@rcc         # agentic refactoring pipeline (optional, independent)

Both plugins live in the same marketplace. Install whichever you need; they coexist and are independently versioned.

How It Works

Agent System Pipeline

The full pipeline runs through 6 stages, each a dedicated skill:

migrate → analyze → brainstorm → plan → apply → review → refactor

Stage	Skill	What it does
Migrate	migrating-agent-systems	Detect maturity (None/Seed/Partial/Established), propose rules refactoring, route to correct chain