How do I know if my subagent is effective?

An effective agent has a specific description with trigger keywords, a system prompt with concrete examples, and tool access restricted to only what is necessary for its domain.

What are Claude Code subagents?

Subagents are specialized AI assistants with their own tools, models, and system prompts that operate in isolated context windows to handle specific tasks without cluttering your main conversation.

When should I use a subagent instead of a direct tool call?

Subagents are best for complex, multi-step tasks, parallel exploration, or processes that generate a high volume of output that would otherwise consume your main context window.

Where are custom subagent configurations stored?

Custom agents are stored as Markdown files within the .claude/agents/ directory for project-specific tasks or ~/.claude/agents/ for global user-level availability.

Managing Subagents

Name: Managing Subagents
Author: thoeltig

bythoeltig

Productivity & Workflow

Optimizes task delegation and subagent configurations to enhance Claude's performance and context preservation.

About

This skill provides a systematic framework for analyzing, evaluating, and creating specialized subagents within the Claude Code environment. It helps developers decide when to delegate tasks to isolated context windows versus using direct tool calls, ensuring that the main conversation remains clean and focused. By utilizing a detailed decision matrix, analysis frameworks, and improvement patterns, this skill empowers users to build high-performance custom agents, refine existing system prompts, and identify outdated configurations that may hinder productivity.

Key Features

Automated evaluation of agent descriptions and system prompts
Step-by-step workflow for creating specialized project-specific subagents
0 GitHub stars
Contextual delegation scoring to determine optimal subagent usage
Performance optimization through model selection and tool restriction analysis
Audit system for identifying outdated Claude Code features and deprecated patterns

Use Cases

Improving a custom 'Security Auditor' agent by refining its trigger keywords and success criteria
Evaluating if a massive codebase exploration task should be offloaded to a specialized Explore agent
Standardizing subagent configurations across a development team to ensure consistent automation workflows

About

Key Features

Automated evaluation of agent descriptions and system prompts
Step-by-step workflow for creating specialized project-specific subagents
0 GitHub stars
Contextual delegation scoring to determine optimal subagent usage
Performance optimization through model selection and tool restriction analysis
Audit system for identifying outdated Claude Code features and deprecated patterns

Use Cases

Improving a custom 'Security Auditor' agent by refining its trigger keywords and success criteria
Evaluating if a massive codebase exploration task should be offloaded to a specialized Explore agent
Standardizing subagent configurations across a development team to ensure consistent automation workflows