agent-architect

Classify knowledge tasks and delegate them to appropriate Claude Code skills via the Skill tool.

16|5|Updated Dec 3, 2025
One-click install
npx skills add https://github.com/wayne930242/Reflexive-Claude-Code --skill agent-architect-wayne930242
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-architect
Source: https://github.com/wayne930242/Reflexive-Claude-Code/tree/main/skills/agent-architect
Command: npx skills add https://github.com/wayne930242/Reflexive-Claude-Code --skill agent-architect-wayne930242

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Claude Code architecture advisor that classifies knowledge and delegates tasks to the appropriate skills, acting as an orchestration gatekeeper rather than executor.

Core Features & Use Cases

  • Classification Framework: Distinguishes LAW vs KNOWLEDGE and routes work to the correct skill.
  • Mandatory Delegation: Always delegates using the Skill tool; never performs work directly.
  • Task Routing: Guides flow for creating/updating CLAUDE.md, rules, hooks, subagents, etc.

Quick Start

Provide a user query; the agent classifies it and immediately delegates to the appropriate skill (e.g., write-skill, write-rules, write-hook).

Frequently Asked Questions about agent-architect

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I route tasks to the right skill in a workflow?

Task routing classifies knowledge work and delegates to appropriate skills based on task type. Agent-architect distinguishes between LAW and KNOWLEDGE tasks, then immediately invokes the correct skill—such as write-skill for skill creation, write-rules for rule updates, or write-hook for hook implementation—ensuring work reaches the right executor.

What's the difference between LAW and KNOWLEDGE task classification?

LAW tasks modify system constraints, rules, hooks, or CLAUDE.md—structural guardrails. KNOWLEDGE tasks create or refine skills, commands, and subagents—functional logic. Classification determines routing: LAW delegates to rule or hook writers; KNOWLEDGE delegates to skill or subagent builders.

Can I delegate multiple workflow steps automatically?

Yes. Mandatory delegation enforces that each step—creating skills, updating CLAUDE.md, writing rules, or spawning subagents—immediately invokes the appropriate skill via the Skill tool. The architecture reviews results before proceeding to the next step, maintaining orchestration control.

When should I use an architecture advisor instead of executing tasks directly?

Use an architecture advisor when workflows involve creating or modifying skills, rules, hooks, subagents, or CLAUDE.md. Direct execution loses routing intelligence and accountability. Delegation ensures each task reaches a specialized handler, reducing errors and maintaining system consistency across orchestrated work.

What happens if a delegated task fails?

The architecture advisor reviews results after delegation and can halt or retry. Mandatory delegation with result review prevents silent failures and allows the orchestrator to diagnose, log, or escalate before proceeding, ensuring workflow reliability.