| name | guid-agent-creation |
| description | Use when creating specialized sub-agents with domain expertise, proper YAML formatting, and clear scope boundaries. Follows Guild system's dynamic specialist creation patterns. |
| model | inherit |
Agent Creation Skill
Pattern Description
What: Creating specialized sub-agents with focused domain expertise and proper Guild system formatting When: Need to create domain-specialist agents for complex workflows or recurring task types Context: Use when existing general-purpose agents lack sufficient domain knowledge for efficient task execution
Guild Agent Creation Framework
Agent Structure Requirements
Mandatory YAML Frontmatter:
---
name: [domain]-specialist
model: inherit
thinking_mode: ultrathink
description: "[One-line summary of specialist expertise]"
---
Key Frontmatter Fields:
- name: Hyphenated, domain-specific (e.g., "frontend-specialist", "api-security-specialist")
- model: inherit (uses user's selected model)
- thinking_mode: ultrathink (enables deep reasoning)
- description: REQUIRED - Clear, concise expertise summary for discovery
Domain Specialist Design Pattern
1. Domain Definition
Clear Scope of Expertise:
- Define specific domain (e.g., "React component optimization", "API security", "database performance")
- Establish boundaries: What's IN vs OUT of scope
- Define collaboration patterns with other domains
2. Context Integration
Project-Specific Intelligence:
- Project Landscape: Architecture, frameworks, conventions currently in use
- Technical Environment: Libraries, tools, deployment infrastructure
- Business Context: Why this domain matters to project goals
- Relevant Knowledge: Available patterns, documentation, existing solutions
3. Agent Content Structure
## Identity
Clear name and description of domain expertise
## Scope
What types of tasks this specialist excels at:
- Primary responsibilities
- Decision-making authority
- Integration boundaries
## Context
Project-specific patterns, tools, and conventions:
- Technology stack in use
- Architectural patterns
- Development workflows
- Quality standards
## Standards
Quality expectations and best practices:
- Code quality criteria
- Testing requirements
- Documentation standards
- Performance considerations
## Integration
How the specialist works with other agents:
- When to collaborate
- When to delegate
- Communication protocols
Creation Guidelines
When to Create Specialists
High Value Scenarios:
- Cross-Domain Complexity: Task spans multiple technical areas requiring deep coordination
- Deep Expertise Needed: Surface-level knowledge insufficient for quality outcomes
- No Existing Agent: Current agents don't cover the required domain effectively
- Recurring Need: Domain appears frequently enough to justify specialization
Avoid Over-Specialization:
- Create generalist-capable specialists who can handle adjacent domains
- Ensure specialists can collaborate rather than work in isolation
- Balance specialist creation with generalist problem-solving capabilities
Agent Creation Process
Step 1: Domain Analysis
- Identify domain boundaries and scope
- Analyze required knowledge and skills
- Determine collaboration needs
- Define success criteria
Step 2: Context Gathering
- Extract project-specific patterns
- Identify relevant tools and technologies
- Document existing conventions
- Map integration points
Step 3: Agent Crafting
- Write clear, focused description
- Define scope and boundaries explicitly
- Embed project context and conventions
- Establish integration patterns
Step 4: Validation
- Verify YAML frontmatter completeness
- Test agent with sample tasks
- Ensure no duplication with existing agents
- Validate collaboration patterns
Critical Rules
✅ ALWAYS Follow These Patterns
- Use hyphenated naming:
[domain]-specialist - Include
thinking_mode: ultrathinkfor deep reasoning - Write specific, expertise-focused descriptions
- Embed actual project context, not generic patterns
- Define clear scope boundaries
❌ NEVER Do These
- Don't create agents without clear domain expertise
- Don't omit the description field (required for discovery)
- Don't reference skills directly (Claude Code discovers them natively)
- Don't create overlapping specialist domains
- Don't use generic descriptions like "handles web development"
Integration with Guild System
Discovery and Usage
- Agents are created during
/guild:setupor dynamically as needed - Guild workflow coordinates agent selection and task delegation
- Agents maintain separate context windows for deep focus
- Collaboration patterns enable complex multi-domain workflows
Relationship to Skills
- Agents: Execution specialists with domain expertise
- Skills: Pattern knowledge and conventions
- Integration: Agents apply skill patterns through contextual understanding, not direct references
Quality Checklist
Before finalizing an agent:
- Has specific, expertise-focused description
- Domain boundaries are clearly defined
- Project context is embedded, not generic
- Collaboration patterns are established
- YAML frontmatter is complete and valid
- No duplication with existing agents
- Success criteria are measurable
Common Pitfalls
❌ Over-Specialization
Problem: Creating agents for trivial or rarely-needed domains Why It Fails: Creates maintenance overhead and fragmentation Better Approach: Create generalist-capable specialists who can handle adjacent domains
❌ Generic Descriptions
Problem: Using vague descriptions like "web development specialist" Why It Fails: Prevents intelligent agent selection and task matching Better Approach: "React component optimization specialist" or "REST API security specialist"
❌ Missing Context
Problem: Creating generic agents without project-specific intelligence Why It Fails: Agents can't apply project conventions effectively Better Approach: Embed actual project patterns, tools, and architectural decisions
❌ Scope Creep
Problem: Agents whose expertise overlaps significantly with existing agents Why It Fails: Creates confusion and inefficient delegation Better Approach: Define clear boundaries and collaboration protocols
Related Resources
Related Skills
- skill-creation: For creating pattern knowledge that agents can apply
- domain-analysis: For identifying specialization opportunities
- project-pattern-discovery: For understanding project context
Related Agents
- general-purpose: For tasks that don't warrant specialization
- guild-coordinator: For orchestrating multi-agent workflows
- domain-specialist: Template for new specialist agents
External Resources
- Guild System Framework - Core architecture principles
- Agent Creation Templates - Standardized agent patterns
- Multi-Agent Workflows - Collaboration patterns