aiwg/agentic/code/frameworks/sdlc-complete/schemas/flows/semantic-agent-discovery.yaml

Cognitive architecture for AI-augmented software development with structured memory, ensemble validation, and closed-loop correction. FAIR-aligned artifacts, 84% cost reduction via human-in-the-loop, standards adopted by 100+ organizations.

# Semantic Agent Discovery Schema
# Based on REF-061 Papazoglou SOA
# Issue: #185

$schema: "https://json-schema.org/draft/2020-12/schema"
$id: "https://aiwg.io/schemas/semantic-agent-discovery/v1"
title: "Semantic Agent Discovery Schema"
description: |
  Schema for service-oriented architecture patterns enabling semantic agent
  discovery based on capabilities rather than fixed registry positions,
  per REF-061 Papazoglou SOA (2003) patterns.

type: object
required:
  - version
  - discovery_config
  - capability_registry

properties:
  version:
    type: string
    pattern: "^\\d+\\.\\d+\\.\\d+$"
    default: "1.0.0"

  discovery_config:
    $ref: "#/$defs/DiscoveryConfig"

  capability_registry:
    $ref: "#/$defs/CapabilityRegistry"

  query_interface:
    $ref: "#/$defs/QueryInterface"

$defs:
  DiscoveryConfig:
    type: object
    description: "Discovery service configuration"
    properties:
      enabled:
        type: boolean
        default: true

      index_path:
        type: string
        default: ".aiwg/agent-discovery/capability-index.json"

      refresh_strategy:
        type: string
        enum:
          - on_startup        # Build index at session start
          - on_demand         # Build when first query made
          - continuous        # Watch for agent changes
        default: on_startup

      embedding_model:
        type: string
        default: "text-embedding-3-small"
        description: "Model for semantic capability matching"

      cache_ttl_seconds:
        type: integer
        default: 3600
        description: "How long to cache discovery results"

      performance_target_ms:
        type: integer
        default: 100
        description: "Target query response time"

  CapabilityRegistry:
    type: object
    description: "Registry of agent capabilities"
    properties:
      sources:
        type: array
        items:
          type: string
        default:
          - ".claude/agents/"
          - "agentic/code/frameworks/sdlc-complete/agents/"
        description: "Paths to scan for agent manifests"

      schema:
        $ref: "#/$defs/AgentCapabilitySchema"

  AgentCapabilitySchema:
    type: object
    description: "Schema for agent capability descriptions"
    required:
      - agent_name
      - capabilities
    properties:
      agent_name:
        type: string

      display_name:
        type: string

      description:
        type: string
        description: "Human-readable agent description"

      domains:
        type: array
        items:
          type: string
        description: "Knowledge domains (security, testing, architecture)"

      operations:
        type: array
        items:
          $ref: "#/$defs/Operation"
        description: "Operations this agent can perform"

      inputs:
        type: array
        items:
          $ref: "#/$defs/ArtifactType"
        description: "Artifact types this agent accepts"

      outputs:
        type: array
        items:
          $ref: "#/$defs/ArtifactType"
        description: "Artifact types this agent produces"

      quality_metrics:
        type: object
        additionalProperties:
          type: number
        description: "Quality scores from performance tracking"

      semantic_embedding:
        type: array
        items:
          type: number
        description: "Vector embedding of capabilities"

  Operation:
    type: object
    required:
      - name
      - description
    properties:
      name:
        type: string
      description:
        type: string
      synonyms:
        type: array
        items:
          type: string
        description: "Alternative ways to describe this operation"
      examples:
        type: array
        items:
          type: string
        description: "Example queries that match this operation"

  ArtifactType:
    type: object
    required:
      - type
    properties:
      type:
        type: string
      schema_ref:
        type: string
      description:
        type: string

  QueryInterface:
    type: object
    description: "Query interface configuration"
    properties:
      query_types:
        type: array
        items:
          $ref: "#/$defs/QueryType"

  QueryType:
    type: object
    properties:
      name:
        type: string
      description:
        type: string
      parameters:
        type: array
        items:
          type: object
          properties:
            name:
              type: string
            type:
              type: string
            required:
              type: boolean
            description:
              type: string

# Standard query types
query_types:
  - name: "semantic"
    description: "Natural language capability query"
    parameters:
      - name: query
        type: string
        required: true
        description: "Natural language description of needed capability"
      - name: top_k
        type: integer
        required: false
        description: "Number of results to return (default: 5)"
    examples:
      - "find agents that can review code for security vulnerabilities"
      - "who can design database schemas"
      - "agent for writing API documentation"

  - name: "domain"
    description: "Query by knowledge domain"
    parameters:
      - name: domain
        type: string
        required: true
        description: "Domain name (security, testing, architecture, etc.)"
      - name: min_quality
        type: number
        required: false
        description: "Minimum quality score (0-1)"
    examples:
      - domain: "security"
      - domain: "testing", min_quality: 0.8

  - name: "operation"
    description: "Query by operation type"
    parameters:
      - name: operation
        type: string
        required: true
        description: "Operation name or description"
      - name: input_type
        type: string
        required: false
        description: "Required input artifact type"
      - name: output_type
        type: string
        required: false
        description: "Required output artifact type"
    examples:
      - operation: "review"
      - operation: "generate", output_type: "TestSuite"

  - name: "artifact_transform"
    description: "Find agents that transform artifact A to B"
    parameters:
      - name: input_type
        type: string
        required: true
      - name: output_type
        type: string
        required: true
    examples:
      - input_type: "PRD", output_type: "SystemDesign"
      - input_type: "Implementation", output_type: "TestSuite"

# Discovery result schema
discovery_result:
  type: object
  required:
    - query
    - results
    - timestamp
  properties:
    query:
      type: object
      description: "Original query"
    results:
      type: array
      items:
        type: object
        properties:
          agent_name:
            type: string
          relevance_score:
            type: number
            minimum: 0
            maximum: 1
          match_explanation:
            type: string
          capabilities_matched:
            type: array
            items:
              type: string
          quality_score:
            type: number
    total_matches:
      type: integer
    query_time_ms:
      type: number
    timestamp:
      type: string
      format: date-time

# Capability index schema
capability_index:
  type: object
  required:
    - version
    - agents
    - last_updated
  properties:
    version:
      type: string
    agents:
      type: array
      items:
        $ref: "#/$defs/IndexedAgent"
    domain_index:
      type: object
      additionalProperties:
        type: array
        items:
          type: string
      description: "Domain -> agent names mapping"
    operation_index:
      type: object
      additionalProperties:
        type: array
        items:
          type: string
      description: "Operation -> agent names mapping"
    input_index:
      type: object
      additionalProperties:
        type: array
        items:
          type: string
      description: "Input type -> agent names mapping"
    output_index:
      type: object
      additionalProperties:
        type: array
        items:
          type: string
      description: "Output type -> agent names mapping"
    last_updated:
      type: string
      format: date-time

  IndexedAgent:
    type: object
    properties:
      name:
        type: string
      manifest_path:
        type: string
      capability_hash:
        type: string
        description: "Hash to detect changes"
      embedding:
        type: array
        items:
          type: number

# Agent protocol
agent_protocol:
  build_index:
    description: "Build capability index from agent manifests"
    triggers:
      - startup
      - manual_request
      - manifest_change_detected
    steps:
      - scan_agent_directories
      - for_each_agent_manifest:
          - parse_capability_metadata
          - extract_domains
          - extract_operations
          - extract_input_output_types
          - generate_semantic_embedding
      - build_domain_index
      - build_operation_index
      - build_artifact_type_indices
      - persist_capability_index

  discover_agents:
    description: "Find agents matching query"
    steps:
      - parse_query_type
      - if_semantic_query:
          - generate_query_embedding
          - compute_similarity_scores
          - rank_by_relevance
      - if_domain_query:
          - lookup_domain_index
          - filter_by_quality
      - if_operation_query:
          - lookup_operation_index
          - filter_by_artifact_types
      - if_artifact_transform_query:
          - intersect_input_output_indices
      - sort_results_by_relevance
      - apply_quality_weighting
      - return_top_k_results

  update_quality_scores:
    description: "Update agent quality scores from performance data"
    triggers:
      - task_completion
      - performance_report_generated
    steps:
      - load_performance_stats
      - update_agent_quality_metrics
      - recalculate_rankings
      - persist_updated_index

# Quality weighting
quality_weighting:
  enabled: true
  factors:
    relevance_score:
      weight: 0.60
      description: "Semantic match to query"
    success_rate:
      weight: 0.25
      description: "Historical task success rate"
    efficiency:
      weight: 0.15
      description: "Token efficiency vs benchmark"

# Pre-computed capability embeddings
standard_capabilities:
  security_review:
    text: "review code for security vulnerabilities, OWASP compliance, threat modeling"
    domains: [security]
    operations: [review, analyze, audit]

  architecture_design:
    text: "design system architecture, create ADRs, evaluate technical approaches"
    domains: [architecture]
    operations: [design, evaluate, document]

  test_generation:
    text: "generate test suites, create test plans, ensure code coverage"
    domains: [testing]
    operations: [generate, create, verify]

  code_review:
    text: "review code quality, identify bugs, suggest improvements"
    domains: [quality]
    operations: [review, analyze, suggest]

  documentation:
    text: "write technical documentation, API docs, user guides"
    domains: [documentation]
    operations: [write, document, explain]

  requirements_analysis:
    text: "analyze requirements, create user stories, define acceptance criteria"
    domains: [requirements]
    operations: [analyze, create, define]

  deployment:
    text: "deploy applications, configure CI/CD, manage infrastructure"
    domains: [devops]
    operations: [deploy, configure, automate]

# CLI integration
cli_commands:
  discover:
    command: "aiwg agent discover <query>"
    description: "Find agents matching natural language query"
    options:
      - name: "--domain"
        description: "Filter by domain"
      - name: "--min-quality"
        description: "Minimum quality score"
      - name: "--top"
        description: "Number of results"

  index:
    command: "aiwg agent index"
    description: "Rebuild capability index"
    options:
      - name: "--force"
        description: "Force full rebuild"

  capabilities:
    command: "aiwg agent capabilities <agent-name>"
    description: "Show agent capabilities"

# Storage
storage:
  index_path: ".aiwg/agent-discovery/capability-index.json"
  embeddings_path: ".aiwg/agent-discovery/embeddings/"
  query_cache_path: ".aiwg/agent-discovery/cache/"
  query_log_path: ".aiwg/logs/discovery-queries.jsonl"

# Performance targets (from REF-061 SOA principles)
research_targets:
  query_response_time: "<100ms"
  index_build_time: "<5s for 100 agents"
  relevance_accuracy: ">85% top-3 precision"

# References
references:
  research:
    - "@.aiwg/research/references/REF-061-papazoglou-soa.md"
    - "@.aiwg/research/gap-analysis.md"
  implementation:
    - "#185"
  related:
    - "@src/extensions/registry.ts"
    - "@src/extensions/types.ts"
    - "@agentic/code/frameworks/sdlc-complete/schemas/flows/uct-agent-selection.yaml"
    - "@agentic/code/frameworks/sdlc-complete/schemas/flows/agent-pubsub.yaml"