What reasoning workflow does this skill recommend?

It promotes a 'Decompose → Solve → Produce → Self-Critique' order, ensuring the agent evaluates its own work only after a full solution has been drafted to improve quality.

Can I limit what tools a custom agent can use?

Yes, the skill guides you on using the 'tools' field in the YAML frontmatter to restrict available capabilities, following the security principle of least privilege.

Why is the agent description field so critical?

The description field serves as the trigger mechanism. It tells the parent Claude instance exactly when to invoke the specialized agent using 'Use this agent when...' patterns and concrete examples.

Is this skill compatible with other AI coding tools?

While optimized for Claude Code, the patterns and structures provided are compatible with other context-aware tools like OpenCode, Cursor, and Gemini CLI.

What is a Claude Code Agent?

A Claude Code agent is an autonomous subprocess spawned via the Task tool to handle complex, multi-step tasks independently with its own isolated context window.

Custom Agent Architect

Name: Custom Agent Architect
Author: NeoLabHQ

byNeoLabHQ

•

542

•

生産性とワークフロー

Streamlines the creation of high-performance Claude Code agents using Anthropic-aligned structures, triggering logic, and optimized system prompts.

The Custom Agent Architect skill provides a comprehensive framework for developing autonomous subprocesses within Claude Code. It ensures every agent is built with professional-grade YAML frontmatter, precise triggering conditions, and sophisticated system prompts that follow proven reasoning patterns. By standardizing the 'Decompose → Solve → Produce → Self-Critique' workflow, this skill enables developers to build reliable, domain-specific AI assistants that trigger accurately based on conversation context and execute complex tasks with higher autonomy and quality.

主な機能

01542 GitHub stars

02Visual UI customization via color coding and model inheritance settings

03Advanced triggering logic using context-aware example blocks and commentary

04Proven system prompt templates with clear behavioral boundaries and edge-case handling

05Standardized YAML frontmatter generation for agent metadata and tool restriction

06Optimized multi-stage reasoning workflows with late-stage self-critique

ユースケース

01Creating proactive documentation assistants that trigger after specific file edits

02Developing domain-specific architects for managing complex refactoring tasks

03Building specialized security or performance auditors for enterprise codebases

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add neolabhq/context-engineering-kit create-agent

For use in Claude.ai and ChatGPT

Download Skill