OSSA Manifest Object

The root object of an OSSA agent manifest. This is the top-level container for all agent configuration.

Field Reference

Metadata Object

Agent metadata for identification and organization.

Metadata Examples

Basic metadata:

metadata:
  name: documentation-agent
  version: 1.2.0
  description: Automated documentation generator for API endpoints

With labels and annotations:

metadata:
  name: security-scanner
  version: 2.0.1
  description: Kubernetes security vulnerability scanner
  labels:
    environment: production
    team: security
    domain: infrastructure
    capability: vulnerability-scanning
  annotations:
    repository: https://github.com/example/security-scanner
    maintainer: security-team@example.com
    documentation: https://docs.example.com/agents/security-scanner

Extensions Object

Framework-specific extensions enable integration with various agent platforms and orchestration frameworks. All extensions are optional.

Extensions Example

apiVersion: ossa/v0.3.x
kind: Agent
metadata:
  name: multi-platform-agent
  version: 1.0.0
spec:
  # ... agent spec ...

extensions:
  kagent:
    kubernetes:
      namespace: agents-production
      labels:
        environment: production
    deployment:
      replicas: 3
      strategy: rolling-update

  buildkit:
    deployment:
      replicas:
        min: 2
        max: 8
    container:
      runtime: docker

  openai_agents:
    enabled: true
    model: gpt-4o
    guardrails:
      enabled: true
      max_tool_calls: 10
      timeout_seconds: 300

Complete Example

apiVersion: ossa/v0.3.x
kind: Agent

metadata:
  name: code-quality-analyzer
  version: 2.1.0
  description: |
    Automated code quality analyzer that reviews pull requests,
    identifies security vulnerabilities, checks coding standards,
    and provides actionable improvement recommendations.
  labels:
    domain: development
    subdomain: quality-assurance
    capability: code-analysis
    team: platform-engineering
    environment: production
  annotations:
    repository: https://github.com/example/code-quality-analyzer
    documentation: https://docs.example.com/agents/code-quality-analyzer
    maintainer: platform-team@example.com
    cost-center: engineering-tools
    last-updated: "2024-11-17"

spec:
  taxonomy:
    domain: development
    subdomain: quality-assurance
    capability: static-analysis

  role: |
    You are a senior code quality analyst specializing in:
    - Security vulnerability detection
    - Performance optimization opportunities
    - Code maintainability assessment
    - Best practices enforcement
    - Documentation quality review

    Analyze code with constructive, actionable feedback.

  llm:
    provider: anthropic
    model: claude-3-5-sonnet-20241022
    temperature: 0.2
    maxTokens: 8192

  tools:
    - type: mcp
      server: filesystem
      capabilities:
        - read_file
        - search_files
    - type: mcp
      server: git
      capabilities:
        - git_diff
        - git_log
        - git_blame
    - type: http
      endpoint: https://api.example.com/static-analysis
      capabilities:
        - run_linter
        - check_complexity

  autonomy:
    level: supervised
    approval_required: false
    allowed_actions:
      - read_code
      - analyze_patterns
      - generate_report
      - post_comment
    blocked_actions:
      - modify_code
      - merge_pr
      - deploy

  constraints:
    cost:
      maxTokensPerDay: 5000000
      maxCostPerDay: 100.0
      currency: USD
    performance:
      maxLatencySeconds: 60
      timeoutSeconds: 300
    resources:
      cpu: "2"
      memory: 4Gi

  observability:
    tracing:
      enabled: true
      exporter: otlp
      endpoint: https://telemetry.example.com/v1/traces
    metrics:
      enabled: true
      exporter: prometheus
      endpoint: https://metrics.example.com
    logging:
      level: info
      format: json

extensions:
  kagent:
    kubernetes:
      namespace: agents-production
      labels:
        cost-center: engineering
    deployment:
      replicas: 2
      strategy: rolling-update

  buildkit:
    deployment:
      replicas:
        min: 1
        max: 5

  openai_agents:
    enabled: false

  anthropic:
    enabled: true
    model: claude-3-5-sonnet-20241022
    max_tokens: 8192
    streaming: true

Validation Rules

apiVersion Validation

• Must match regex: ^ossa/v(0\.2\.[2-3]|1)(\.0)?(-[a-zA-Z0-9]+)?$
• Valid examples: ossa/v1, ossa/v0.3.x, ossa/v0.3.0, ossa/v1.0
• Invalid examples: v1, ossa/v2, ossa/0.3.0

kind Validation

• Must be exactly Agent (case-sensitive)
• No other resource types currently supported

metadata.name Validation

• DNS-1123 subdomain format
• Lowercase letters, numbers, and hyphens only
• Must start and end with alphanumeric character
• Maximum 253 characters
• Regex: ^[a-z0-9]([-a-z0-9]*[a-z0-9])?$

Valid names:

• my-agent
• agent-001
• data-processor-v2

Invalid names:

• MyAgent (uppercase)
• -agent (starts with hyphen)
• agent_01 (underscore not allowed)

metadata.version Validation

• Must follow semver 2.0.0 specification
• Format: MAJOR.MINOR.PATCH[-PRERELEASE][+BUILDMETADATA]
• Examples: 1.0.0, 2.1.3-beta.1, 1.0.0+20240101

Related Objects

• Agent Spec - The spec object
• LLM Configuration - LLM settings in spec.llm
• Tools - Tool definitions in spec.tools
• Autonomy - Autonomy settings in spec.autonomy
• Constraints - Constraints in spec.constraints
• Observability - Observability in spec.observability
• Extensions - Framework-specific extensions

JSON Schema

The complete JSON Schema is available at:

https://openstandardagents.org/schemas/v0.3.1/agent.json

Use it for validation and editor autocomplete:

{
  "$schema": "https://openstandardagents.org/schemas/v0.3.1/agent.json",
  "apiVersion": "ossa/v0.3.x",
  "kind": "Agent",
  "metadata": { ... },
  "spec": { ... }
}