aiwg/agentic/code/frameworks/sdlc-complete/schemas/research/quality-dimensions.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.

# Quality Dimensions Schema
# Based on REF-015 Self-Refine Research
# Issues: #147, #148, #149

$schema: "https://json-schema.org/draft/2020-12/schema"
$id: "https://aiwg.io/schemas/quality-dimensions/v1"
title: "Multi-Dimensional Quality Scoring Schema"
description: |
  Schema for tracking quality across multiple dimensions per artifact,
  measuring feedback accuracy, and early stopping per REF-015.

type: object
required:
  - version
  - dimensions

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

  dimensions:
    type: array
    items:
      $ref: "#/$defs/QualityDimension"

  scoring:
    $ref: "#/$defs/ScoringConfig"

  feedback:
    $ref: "#/$defs/FeedbackConfig"

  early_stopping:
    $ref: "#/$defs/EarlyStoppingConfig"

$defs:
  QualityDimension:
    type: object
    required:
      - id
      - name
      - description
    properties:
      id:
        type: string
        pattern: "^[a-z][a-z0-9_]*$"
        description: "Dimension identifier"

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

      description:
        type: string
        description: "What this dimension measures"

      weight:
        type: number
        minimum: 0
        maximum: 1
        default: 1.0
        description: "Weight in aggregate score"

      artifact_types:
        type: array
        items:
          type: string
        description: "Artifact types this applies to (empty = all)"

      evaluation_prompt:
        type: string
        description: "Prompt template for LLM-based scoring"

      automated_checks:
        type: array
        items:
          type: string
        description: "Automated validation checks"

  ScoringConfig:
    type: object
    properties:
      scale:
        type: object
        properties:
          min:
            type: integer
            default: 0
          max:
            type: integer
            default: 100

      thresholds:
        type: object
        properties:
          excellent:
            type: integer
            default: 90
          good:
            type: integer
            default: 75
          acceptable:
            type: integer
            default: 60
          poor:
            type: integer
            default: 40

      aggregate_method:
        type: string
        enum: [weighted_average, minimum, geometric_mean]
        default: weighted_average

      storage_path:
        type: string
        default: ".aiwg/.quality/"

  FeedbackConfig:
    type: object
    description: "Feedback accuracy tracking per REF-015"
    properties:
      ab_testing:
        type: object
        properties:
          enabled:
            type: boolean
            default: true
          sample_rate:
            type: number
            default: 0.2
            description: "Fraction of artifacts in test group"

      accuracy_metrics:
        type: object
        properties:
          track_quality_delta:
            type: boolean
            default: true
          track_false_positives:
            type: boolean
            default: true
          track_false_negatives:
            type: boolean
            default: true

      feedback_sources:
        type: array
        items:
          type: string
          enum:
            - self_critique
            - peer_review
            - human_review
            - automated_validation

      storage_path:
        type: string
        default: ".aiwg/.telemetry/feedback/"

  EarlyStoppingConfig:
    type: object
    description: "Early stopping configuration per REF-015"
    properties:
      enabled:
        type: boolean
        default: true

      conditions:
        type: object
        properties:
          confidence_threshold:
            type: integer
            default: 90
            description: "Min aggregate quality score"
          all_verifications_pass:
            type: boolean
            default: true
          min_iterations:
            type: integer
            default: 1
            description: "Minimum iterations before early stop"
          max_iterations:
            type: integer
            default: 10
            description: "Hard limit if confidence never reached"

      verifications:
        type: array
        items:
          type: string
        default:
          - tests_pass
          - lint_clean
          - type_check_clean
          - security_scan_clean

      user_override:
        type: boolean
        default: true
        description: "Allow user to continue despite high confidence"

      telemetry:
        type: object
        properties:
          track_early_stops:
            type: boolean
            default: true
          track_iterations_saved:
            type: boolean
            default: true

# Default quality dimensions
default_dimensions:
  - id: correctness
    name: "Correctness"
    description: "Does the artifact correctly implement requirements?"
    weight: 1.0
    evaluation_prompt: |
      Evaluate correctness of this {artifact_type}:
      - Does it meet stated requirements?
      - Are there logical errors?
      - Does it handle edge cases?
      Score 0-100 with justification.
    automated_checks:
      - tests_pass
      - no_runtime_errors

  - id: completeness
    name: "Completeness"
    description: "Does the artifact cover all required aspects?"
    weight: 0.9
    evaluation_prompt: |
      Evaluate completeness of this {artifact_type}:
      - Are all requirements addressed?
      - Are there missing features or sections?
      - Is documentation complete?
      Score 0-100 with justification.
    automated_checks:
      - no_todo_markers
      - all_sections_present

  - id: efficiency
    name: "Efficiency"
    description: "Is the artifact optimally implemented?"
    weight: 0.7
    artifact_types:
      - source_code
      - test
      - script
    evaluation_prompt: |
      Evaluate efficiency of this code:
      - Are there performance issues?
      - Is there unnecessary complexity?
      - Could this be optimized?
      Score 0-100 with justification.
    automated_checks:
      - no_obvious_inefficiencies
      - complexity_within_limits

  - id: readability
    name: "Readability"
    description: "Is the artifact easy to understand?"
    weight: 0.8
    evaluation_prompt: |
      Evaluate readability of this {artifact_type}:
      - Is the structure clear?
      - Are names descriptive?
      - Is formatting consistent?
      Score 0-100 with justification.
    automated_checks:
      - lint_clean
      - formatting_consistent

  - id: maintainability
    name: "Maintainability"
    description: "Can the artifact be easily modified and extended?"
    weight: 0.8
    evaluation_prompt: |
      Evaluate maintainability of this {artifact_type}:
      - Is it modular and well-organized?
      - Are dependencies clear?
      - Would changes be straightforward?
      Score 0-100 with justification.
    automated_checks:
      - low_coupling
      - high_cohesion

  - id: security
    name: "Security"
    description: "Does the artifact follow security best practices?"
    weight: 1.0
    artifact_types:
      - source_code
      - configuration
      - script
    evaluation_prompt: |
      Evaluate security of this {artifact_type}:
      - Are there vulnerabilities (OWASP)?
      - Is input validated?
      - Are secrets properly handled?
      Score 0-100 with justification.
    automated_checks:
      - security_scan_clean
      - no_hardcoded_secrets

# Quality profile schema (stored per artifact)
quality_profile:
  artifact_path:
    type: string
  artifact_type:
    type: string
  timestamp:
    type: string
    format: date-time
  iteration:
    type: integer
  scores:
    type: object
    description: "Dimension ID -> score (0-100)"
  aggregate_score:
    type: number
  confidence:
    type: number
    description: "Overall confidence level"
  feedback:
    type: array
    items:
      type: object
      properties:
        dimension:
          type: string
        suggestion:
          type: string
        priority:
          type: string
          enum: [high, medium, low]
  history:
    type: array
    description: "Previous quality profiles for trend tracking"

# Feedback accuracy tracking
feedback_accuracy:
  feedback_id:
    type: string
  source:
    type: string
    enum: [self_critique, peer_review, human_review, automated]
  artifact_path:
    type: string
  dimension:
    type: string
  suggestion:
    type: string
  applied:
    type: boolean
  quality_before:
    type: number
  quality_after:
    type: number
  delta:
    type: number
  was_helpful:
    type: boolean
    description: "delta > 0"

# Agent protocol
agent_protocol:
  quality_scoring:
    description: "Score artifact on all dimensions"
    steps:
      - identify_artifact_type
      - select_applicable_dimensions
      - run_automated_checks
      - generate_llm_scores
      - calculate_aggregate
      - generate_feedback
      - store_profile

  feedback_validation:
    description: "Validate feedback effectiveness"
    steps:
      - record_pre_quality
      - apply_feedback
      - record_post_quality
      - calculate_delta
      - update_accuracy_metrics

  early_stop_check:
    description: "Check if iteration can stop early"
    steps:
      - get_latest_quality_profile
      - check_confidence_threshold
      - run_verification_checks
      - decide_stop_or_continue
      - log_decision

# Telemetry
telemetry:
  metrics:
    - average_quality_by_dimension
    - quality_trend_over_iterations
    - feedback_accuracy_rate
    - false_positive_rate
    - early_stop_frequency
    - iterations_saved

  alerts:
    - quality_degradation: "aggregate_score drops >10 points"
    - low_feedback_accuracy: "<50% feedback helpful"
    - no_early_stops: "0 early stops in 10 iterations"

# References
references:
  research:
    - "@.aiwg/research/findings/REF-015-self-refine.md"
  implementation:
    - "#147"  # Multi-dimensional quality
    - "#148"  # Feedback accuracy
    - "#149"  # Early stopping
  related:
    - "@src/quality/scorer.ts"
    - "@src/quality/feedback-validator.ts"
    - "@src/iteration/early-stop.ts"