custom-agent

name	custom-agent
description	["Describe when this subagent should be invoked. Be specific and action-oriented. Use keywords like PROACTIVELY or MUST BE USED if appropriate."]
tools	Read, Write, Edit, Grep, Glob, Bash
model	inherit

[Define the subagent's role, expertise, and purpose here]

Role

[Describe what this subagent specializes in and what it should accomplish]

Process

When invoked:

1. [First step name]

   • [Action or consideration]
   • [Action or consideration]

2. [Second step name]

   • [Action or consideration]
   • [Action or consideration]

3. [Third step name]

   • [Action or consideration]
   • [Action or consideration]

[Add more steps as needed]

Guidelines

• [Key principle or constraint]
• [Important consideration]
• [Best practice to follow]
• [What to avoid]

Output Format

[Describe how the subagent should structure its responses]

[Optional: Include examples of expected output]

Best Practices

• [Practice 1]
• [Practice 2]
• [Practice 3]

---

Instructions for Customizing This Template

1. Update frontmatter:

   • Change name to a descriptive, hyphenated name (e.g., api-integration-tester)
   • Write a clear description explaining when to invoke this subagent
   • Adjust tools list to only necessary tools (or omit to inherit all)
   • Set model if needed (sonnet, opus, haiku, or inherit)

2. Define the role:

   • Be specific about the subagent's expertise and purpose
   • Include the domain, technology, or workflow focus
   • Set clear expectations for what it can accomplish

3. Document the process:

   • Break down the workflow into clear steps
   • Use action-oriented language
   • Include decision points and considerations
   • Add error handling or edge cases

4. Establish guidelines:

   • Define principles the subagent should follow
   • List important constraints or requirements
   • Note what the subagent should avoid
   • Include best practices specific to this domain

5. Specify output format:

   • Describe how results should be structured
   • Include templates or examples if helpful
   • Define what constitutes successful completion

6. Test and iterate:

   • Test with realistic tasks
   • Refine description for better automatic invocation
   • Adjust process based on actual usage
   • Update guidelines as patterns emerge

Examples of Effective Subagents

Narrow, Focused Subagents

• Security vulnerability scanner (focused on one concern)
• API documentation generator (specific output type)
• Database migration creator (specific workflow)

Domain-Specific Subagents

• React component builder (framework-specific)
• AWS infrastructure manager (platform-specific)
• GraphQL schema designer (technology-specific)

Workflow-Specific Subagents

• Pre-commit validator (workflow stage)
• Release note generator (process step)
• Performance benchmark runner (specific task)

Choose clear boundaries and a single responsibility for best results.