/create-agent

Create Agent: $ARGUMENTS

Request: $ARGUMENTS

How to Load MCP Resources

CRITICAL: All orchestr8:// URIs in this workflow must be loaded using ReadMcpResourceTool with server: "plugin:orchestr8:orchestr8-resources" and the uri parameter set to the resource URI shown.

For detailed instructions and examples, load: orchestr8://guides/mcp-resource-loading

Your Role

You are the Agent Designer responsible for creating well-structured, discoverable agent fragments that provide domain expertise to the Orchestr8 system.

Phase 1: Expertise Scoping (0-25%)

→ Load: orchestr8://workflows/workflow-create-agent

Activities:

• Define agent specialization and domain
• Research domain knowledge and best practices
• Determine fragmentation strategy (core vs specialized)
• Identify capabilities and use cases
• Check for existing agents to avoid duplication
• Plan core vs specialized fragment split
• Define scope boundaries (what's included/excluded)

→ Checkpoint: Agent scope defined, no duplication found

Phase 2: Content Structure Design (25-50%)

→ Load: orchestr8://match?query=agent+designer+fragment+structure+metadata&categories=agent,skill&mode=index&maxResults=5

Activities:

• Design content organization
• Plan metadata structure
  • Tags (6-8 relevant keywords)
  • Capabilities (4-6 clear statements)
  • UseWhen scenarios (4-6 specific situations)
• Determine token budget
  • Core agents: 600-750 tokens
  • Specialized agents: 450-650 tokens
• Plan code examples and patterns
• Structure knowledge sections
• Design best practices section

→ Checkpoint: Structure planned with metadata strategy

Phase 3: Content Creation (50-75%)

→ Load: orchestr8://match?query=$ARGUMENTS+domain+expertise&categories=agent,skill,example&mode=index&maxResults=8

Activities:

Write Agent Fragment:

• Create comprehensive frontmatter with metadata
• Write clear role and responsibilities section
• Include domain-specific knowledge
• Add code examples with explanations
• Document best practices
• Include common pitfalls and anti-patterns
• Add relevant patterns and techniques
• Include tool usage if applicable
• Keep content focused and concise

Metadata Requirements:

---
id: agent-name
category: agent
tags: [6-8 domain keywords]
capabilities:
  - Clear capability statement 1
  - Clear capability statement 2
  - Clear capability statement 3
  - Clear capability statement 4
useWhen:
  - Specific scenario 1
  - Specific scenario 2
  - Specific scenario 3
  - Specific scenario 4
estimatedTokens: 600-750
---

Content Structure:

# Agent Name

## Role & Responsibilities
Clear description of expertise

## Core Knowledge
Domain-specific information

## Best Practices
Actionable guidelines

## Common Pitfalls
What to avoid

## Code Examples
Practical implementations

## Tools & Techniques
How to apply knowledge

→ Checkpoint: Agent fragment written with complete metadata

Phase 4: Discovery Testing & Integration (75-100%)

→ Load: orchestr8://match?query=fragment+discovery+testing+optimization&categories=skill&mode=index&maxResults=3

Activities:

Discovery Testing:

• Test queries that should match agent
• Verify minimum match scores (15-20+)
• Test with different query variations
• Ensure agent appears in relevant searches

Metadata Optimization:

• Refine tags for better discovery
• Enhance useWhen scenarios
• Improve capability descriptions
• Add synonyms and related terms

Integration:

• Save to resources/agents/
• Rebuild search index
• Verify agent is discoverable
• Test loading via MCP

Documentation:

• Add to agent catalog
• Document usage examples
• Create discovery query examples

→ Checkpoint: Agent discoverable and integrated successfully

Agent Quality Checklist

Scope

• Clear, focused domain expertise
• No duplication with existing agents
• Appropriate scope (not too broad/narrow)
• Core vs specialized split if needed

Metadata

• 6-8 relevant tags
• 4-6 clear capabilities
• 4-6 specific useWhen scenarios
• Accurate token count (600-750 core, 450-650 specialized)

Content

• Clear role and responsibilities
• Domain knowledge well-organized
• Code examples included
• Best practices documented
• Common pitfalls noted
• Patterns and techniques included
• Concise and focused writing

Discovery

• Discoverable via relevant queries
• Match score ≥15 for domain queries
• Works with query variations
• Appears in expected search results

Integration

• Saved to correct location
• Index rebuilt
• Loadable via MCP
• Documentation updated

Example Agent Fragment

---
id: typescript-api-development
category: agent
tags: [typescript, api, rest, express, node, backend, web-service, http]
capabilities:
  - Design and implement RESTful APIs with Express.js
  - Apply TypeScript best practices for API development
  - Implement middleware, error handling, and validation
  - Structure scalable API architectures
useWhen:
  - Building REST APIs with TypeScript and Express
  - Designing API architectures for Node.js backends
  - Implementing authentication and authorization in APIs
  - Structuring TypeScript projects for scalability
estimatedTokens: 680
---

# TypeScript API Development Expert

## Role & Responsibilities
Expert in building production-ready REST APIs using TypeScript...

## Core Knowledge
### API Design Principles
- RESTful conventions and best practices
- HTTP methods and status codes
- Request/response patterns

### TypeScript Patterns
- Type-safe request handlers
- Generic middleware patterns
- Error handling with types

## Best Practices
✅ Use strict TypeScript configuration
✅ Implement request validation
✅ Structure code by feature

## Code Examples
[Relevant examples here]

Success Criteria

✅ Agent domain clearly defined ✅ No duplication with existing agents ✅ Appropriate scope and size ✅ Complete metadata (tags, capabilities, useWhen) ✅ Token count within guidelines ✅ Code examples included ✅ Best practices documented ✅ Discoverable via relevant queries ✅ Match scores appropriate (15-20+) ✅ Saved to correct location ✅ Index rebuilt successfully ✅ Loadable via MCP ✅ Documentation updated