When should I use the 'Override' vs 'Append' system prompt architecture?

Use 'Override' when building a standalone product where you want full control over behavior. Use 'Append' when you want to extend the existing capabilities of Claude Code while maintaining its core features.

Can this skill help with multi-agent orchestration?

Yes, it provides deployment form guidance and configuration patterns specifically for building and deploying multi-agent workflows.

Does the Custom Agent Design skill support custom tools?

Yes, it includes specific steps for configuring tool access, including whitelisting/blacklisting standard tools and integrating custom tools via MCP (Model Context Protocol) servers.

How do I choose between Haiku, Sonnet, and Opus for my custom agent?

The skill provides a selection matrix: use Haiku for fast/cheap transformations, Sonnet for balanced performance, and Opus for tasks requiring complex reasoning and highest quality.

Custom Agent Architect

Name: Custom Agent Architect
Author: melodic-software

bymelodic-software

•

数据科学与机器学习

Streamlines the design and architecture of domain-specific AI agents using Claude Agent SDK patterns.

关于

The Custom Agent Design skill provides a structured framework for building specialized, domain-specific AI agents from the ground up. It guides developers through the entire lifecycle of agent creation—from defining a singular purpose and selecting the optimal Claude model to architecting system prompts and configuring granular tool access. Whether you need a simple automation script or a complex multi-agent system, this skill ensures best practices in governance, deployment strategy, and SDK implementation are followed to create robust, reliable AI assistants.

主要功能

System prompt design patterns for Override vs. Append models
1 GitHub stars
Granular tool access and MCP server configuration guidance
Deployment strategy selection for APIs, REPLs, and Streams
Step-by-step agent architecture framework
Security governance implementation with custom tool hooks

使用场景

Building secure, governed AI tools for enterprise environments
Converting generic AI workflows into specialized domain-expert agents
Designing multi-agent systems for complex task orchestration

关于

主要功能

System prompt design patterns for Override vs. Append models
1 GitHub stars
Granular tool access and MCP server configuration guidance
Deployment strategy selection for APIs, REPLs, and Streams
Step-by-step agent architecture framework
Security governance implementation with custom tool hooks

使用场景

Building secure, governed AI tools for enterprise environments
Converting generic AI workflows into specialized domain-expert agents
Designing multi-agent systems for complex task orchestration