agent.define Skill

agent.define Skill

Validates and registers agent manifests for the Betty Framework.

Purpose

The agent.define skill is the Layer 2 (Reasoning Layer) equivalent of skill.define. It validates agent manifests (agent.yaml) for schema compliance, verifies skill references, and updates the central Agent Registry.

Capabilities

• Validate agent manifest structure and required fields
• Verify agent name and version formats
• Validate reasoning mode enum values
• Check that all referenced skills exist in skill registry
• Ensure capabilities and skills lists are non-empty
• Validate status lifecycle values
• Register valid agents in /registry/agents.json
• Update existing agent entries with new versions

Usage

Command Line

python skills/agent.define/agent_define.py <path_to_agent.yaml>

Arguments

Argument	Type	Required	Description
manifest_path	string	Yes	Path to the agent.yaml file to validate

Exit Codes

• 0: Validation succeeded and agent was registered
• 1: Validation failed or registration error

Validation Rules

Required Fields

All agent manifests must include:

Field	Type	Validation
name	string	Must match ^[a-z][a-z0-9._-]*$
version	string	Must follow semantic versioning
description	string	Non-empty string (1-200 chars recommended)
capabilities	array[string]	Must contain at least one item
skills_available	array[string]	Must contain at least one item, all skills must exist in registry
reasoning_mode	enum	Must be iterative or oneshot

Optional Fields

Field	Type	Default	Validation
status	enum	draft	Must be draft, active, deprecated, or archived
context_requirements	object	{}	Any valid object
workflow_pattern	string	null	Any string
example_task	string	null	Any string
error_handling	object	{}	Any valid object
output	object	{}	Any valid object
tags	array[string]	[]	Array of strings
dependencies	array[string]	[]	Array of strings

Name Format

Agent names must:

• Start with a lowercase letter
• Contain only lowercase letters, numbers, dots, hyphens, underscores
• Follow pattern: <domain>.<action>

Valid: api.designer, compliance.checker, data-migrator Invalid: ApiDesigner, 1agent, agent_name

Version Format

Versions must follow semantic versioning: MAJOR.MINOR.PATCH[-prerelease]

Valid: 0.1.0, 1.0.0, 2.3.1-beta, 1.0.0-rc.1 Invalid: 1.0, v1.0.0, 1.0.0.0

Reasoning Mode

Must be one of:

• iterative: Agent can retry with feedback and refine based on errors
• oneshot: Agent executes once without retry

Skills Validation

All skills in skills_available must exist in the skill registry (/registry/skills.json).

Response Format

Success Response

{
  "ok": true,
  "status": "success",
  "errors": [],
  "path": "agents/api.designer/agent.yaml",
  "details": {
    "valid": true,
    "errors": [],
    "path": "agents/api.designer/agent.yaml",
    "manifest": {
      "name": "api.designer",
      "version": "0.1.0",
      "description": "Design RESTful APIs...",
      "capabilities": [...],
      "skills_available": [...],
      "reasoning_mode": "iterative"
    },
    "status": "registered",
    "registry_updated": true
  }
}

Failure Response

{
  "ok": false,
  "status": "failed",
  "errors": [
    "Missing required fields: capabilities, skills_available",
    "Invalid reasoning_mode: 'hybrid'. Must be one of: iterative, oneshot"
  ],
  "path": "agents/bad.agent/agent.yaml",
  "details": {
    "valid": false,
    "errors": [
      "Missing required fields: capabilities, skills_available",
      "Invalid reasoning_mode: 'hybrid'. Must be one of: iterative, oneshot"
    ],
    "path": "agents/bad.agent/agent.yaml"
  }
}

Examples

Example 1: Validate Iterative Agent

Agent Manifest (agents/api.designer/agent.yaml):

name: api.designer
version: 0.1.0
description: "Design RESTful APIs following enterprise guidelines"

capabilities:
  - Design RESTful APIs from requirements
  - Apply Zalando guidelines automatically
  - Generate OpenAPI 3.1 specs
  - Iteratively refine based on validation feedback

skills_available:
  - api.define
  - api.validate
  - api.generate-models

reasoning_mode: iterative

status: draft

tags:
  - api
  - design
  - openapi

Command:

python skills/agent.define/agent_define.py agents/api.designer/agent.yaml

Output:

{
  "ok": true,
  "status": "success",
  "errors": [],
  "path": "agents/api.designer/agent.yaml",
  "details": {
    "valid": true,
    "status": "registered",
    "registry_updated": true
  }
}

Example 2: Validation Errors

Agent Manifest (agents/bad.agent/agent.yaml):

name: BadAgent  # Invalid: must be lowercase
version: 1.0   # Invalid: must be semver
description: "Test agent"
capabilities: []  # Invalid: must have at least one
skills_available:
  - nonexistent.skill  # Invalid: skill doesn't exist
reasoning_mode: hybrid  # Invalid: must be iterative or oneshot

Command:

python skills/agent.define/agent_define.py agents/bad.agent/agent.yaml

Output:

{
  "ok": false,
  "status": "failed",
  "errors": [
    "Invalid name: Invalid agent name: 'BadAgent'. Must start with lowercase letter...",
    "Invalid version: Invalid version: '1.0'. Must follow semantic versioning...",
    "Invalid reasoning_mode: Invalid reasoning_mode: 'hybrid'. Must be one of: iterative, oneshot",
    "capabilities must contain at least one item",
    "Skills not found in registry: nonexistent.skill"
  ],
  "path": "agents/bad.agent/agent.yaml"
}

Example 3: Oneshot Agent

Agent Manifest (agents/api.analyzer/agent.yaml):

name: api.analyzer
version: 0.1.0
description: "Analyze API specifications for compatibility"

capabilities:
  - Detect breaking changes between API versions
  - Generate compatibility reports
  - Suggest migration paths

skills_available:
  - api.compatibility

reasoning_mode: oneshot

output:
  success:
    - Compatibility report
    - Breaking changes list
  failure:
    - Error analysis

status: active

tags:
  - api
  - analysis
  - compatibility

Command:

python skills/agent.define/agent_define.py agents/api.analyzer/agent.yaml

Result: Agent validated and registered successfully.

Integration

With Registry

The skill automatically updates /registry/agents.json:

{
  "registry_version": "1.0.0",
  "generated_at": "2025-10-23T10:30:00Z",
  "agents": [
    {
      "name": "api.designer",
      "version": "0.1.0",
      "description": "Design RESTful APIs following enterprise guidelines",
      "reasoning_mode": "iterative",
      "skills_available": ["api.define", "api.validate", "api.generate-models"],
      "capabilities": ["Design RESTful APIs from requirements", ...],
      "status": "draft",
      "tags": ["api", "design", "openapi"],
      "dependencies": []
    }
  ]
}

With Other Skills

• Depends on: skill.define (for skill registry validation)
• Used by: Future command.define skill (to register commands that invoke agents)
• Complements: workflow.compose (agents orchestrate skills; workflows execute fixed sequences)

Common Errors

Missing Skills in Registry

Error:

Skills not found in registry: api.nonexistent, data.missing

Solution: Ensure all skills in skills_available are registered in /registry/skills.json. Check skill names for typos.

Invalid Reasoning Mode

Error:

Invalid reasoning_mode: 'hybrid'. Must be one of: iterative, oneshot

Solution: Use iterative for agents that retry with feedback, or oneshot for deterministic execution.

Empty Capabilities

Error:

capabilities must contain at least one item

Solution: Add at least one capability string describing what the agent can do.

Invalid Name Format

Error:

Invalid agent name: 'API-Designer'. Must start with lowercase letter...

Solution: Use lowercase names following pattern <domain>.<action> (e.g., api.designer).

Development

Testing

Create test agent manifests in /agents/test/:

# Create test directory
mkdir -p agents/test.agent

# Create minimal test manifest
cat > agents/test.agent/agent.yaml << EOF
name: test.agent
version: 0.1.0
description: "Test agent"
capabilities:
  - Test capability
skills_available:
  - skill.define
reasoning_mode: oneshot
status: draft
EOF

# Validate
python skills/agent.define/agent_define.py agents/test.agent/agent.yaml

Registry Location

• Skill registry: /registry/skills.json (read for validation)
• Agent registry: /registry/agents.json (updated by this skill)

See Also

• Agent Schema Reference - Complete field specifications
• Betty Architecture - Five-layer architecture overview
• Agent Implementation Plan - Implementation details
• /agents/README.md - Agent directory documentation