aiox-core 5.0.3 → 5.0.4

package/pro/squads/squad-creator-pro/test-cases/qa-after-creation/qualification-report.yaml

@@ -1,97 +1,97 @@
-# Qualification Report - qa-after-creation
-# Generated: 2026-02-11
-# Workflow: wf-model-tier-qualification v2.0
-
-qualification_report:
-  task_name: "qa-after-creation"
-  test_date: "2026-02-11"
-  workflow_version: "2.0"
-
-  baseline:
-    model: "opus"
-    final_score: 8.32
-    result: "APPROVED"
-    type_detection: "EXPERT (95% confidence)"
-    methodology: "Full 6-phase validation including type detection"
-    tokens: 113793
-    tool_uses: 39
-    duration_ms: 206004
-
-  rounds:
-    - round: 1
-      model: "haiku"
-      final_score: 9.9
-      result: "APPROVED"
-      type_detection: "N/A (not executed)"
-      methodology: "5-phase quick validation (simplified)"
-      tokens: 59289
-      tool_uses: 26
-      duration_ms: 170335
-      quality_vs_baseline: "119%"
-      gaps:
-        - "Haiku gave HIGHER score than Opus (9.9 vs 8.32)"
-        - "Haiku did NOT do type detection (Opus did EXPERT 95%)"
-        - "Different validation methodology applied"
-        - "Haiku used simplified checklist-based scoring"
-        - "Opus used full validate-squad integration"
-
-  decision:
-    veto_triggered: false
-    veto_id: null
-    final: "HAIKU_QUALIFIED_WITH_CAVEAT"
-    recommended_tier: "haiku"
-    quality_achieved: "119%"
-    threshold_required: "90%"
-
-    caveats:
-      - "Haiku score inflation (+19% vs Opus)"
-      - "Different validation depth (Haiku simplified)"
-      - "Type detection skipped in Haiku approach"
-
-    rationale: |
-      Both models produced APPROVED result - the final decision matches.
-      Score difference (9.9 vs 8.32) is due to methodology difference:
-      - Opus integrated full validate-squad (more rigorous)
-      - Haiku used simplified checklist approach (faster, less deep)
-
-      For QA-after-creation, the key outcome is APPROVED/REJECTED.
-      Both reached same conclusion → Haiku is viable for this task.
-
-  cost_analysis:
-    opus_per_run: "$0.68"
-    haiku_per_run: "$0.012"
-    savings_if_haiku: "98.2%"
-    savings_applicable: true
-    reason: "Same final decision (APPROVED). Score difference acceptable for QA pass/fail."
-
-  methodology_note: |
-    qa-after-creation is a GATEKEEPER task:
-    - Primary purpose: Pass/Fail decision after component creation
-    - Secondary purpose: Quality score for tracking
-
-    For pass/fail: HAIKU MATCHES OPUS ✓
-    For score precision: HAIKU INFLATES (+19%)
-
-    Recommendation: Use Haiku for QA gating decisions.
-    If precise score needed, use Opus or add calibration.
-
-  learning:
-    pattern_identified: |
-      GATEKEEPER tasks (binary pass/fail) are HAIKU-ELIGIBLE
-      even if scores differ, as long as final decision matches.
-
-    similar_tasks:
-      - "qa-after-creation (qualified)"
-      - "validate-extraction (likely qualified)"
-      - "security-scan (likely qualified)"
-
-    different_from:
-      - "validate-squad (requires type detection)"
-      - "pv-modernization-score (requires interpretation)"
-
-metadata:
-  validator: "@pedro-valerio"
-  note: "Output files not created by agents (I/O issue). Report based on completion notifications."
-  tokens_opus: 113793
-  tokens_haiku: 59289
-  total_cost: "$0.69"
+# Qualification Report - qa-after-creation
+# Generated: 2026-02-11
+# Workflow: wf-model-tier-qualification v2.0
+
+qualification_report:
+  task_name: "qa-after-creation"
+  test_date: "2026-02-11"
+  workflow_version: "2.0"
+
+  baseline:
+    model: "opus"
+    final_score: 8.32
+    result: "APPROVED"
+    type_detection: "EXPERT (95% confidence)"
+    methodology: "Full 6-phase validation including type detection"
+    tokens: 113793
+    tool_uses: 39
+    duration_ms: 206004
+
+  rounds:
+    - round: 1
+      model: "haiku"
+      final_score: 9.9
+      result: "APPROVED"
+      type_detection: "N/A (not executed)"
+      methodology: "5-phase quick validation (simplified)"
+      tokens: 59289
+      tool_uses: 26
+      duration_ms: 170335
+      quality_vs_baseline: "119%"
+      gaps:
+        - "Haiku gave HIGHER score than Opus (9.9 vs 8.32)"
+        - "Haiku did NOT do type detection (Opus did EXPERT 95%)"
+        - "Different validation methodology applied"
+        - "Haiku used simplified checklist-based scoring"
+        - "Opus used full validate-squad integration"
+
+  decision:
+    veto_triggered: false
+    veto_id: null
+    final: "HAIKU_QUALIFIED_WITH_CAVEAT"
+    recommended_tier: "haiku"
+    quality_achieved: "119%"
+    threshold_required: "90%"
+
+    caveats:
+      - "Haiku score inflation (+19% vs Opus)"
+      - "Different validation depth (Haiku simplified)"
+      - "Type detection skipped in Haiku approach"
+
+    rationale: |
+      Both models produced APPROVED result - the final decision matches.
+      Score difference (9.9 vs 8.32) is due to methodology difference:
+      - Opus integrated full validate-squad (more rigorous)
+      - Haiku used simplified checklist approach (faster, less deep)
+
+      For QA-after-creation, the key outcome is APPROVED/REJECTED.
+      Both reached same conclusion → Haiku is viable for this task.
+
+  cost_analysis:
+    opus_per_run: "$0.68"
+    haiku_per_run: "$0.012"
+    savings_if_haiku: "98.2%"
+    savings_applicable: true
+    reason: "Same final decision (APPROVED). Score difference acceptable for QA pass/fail."
+
+  methodology_note: |
+    qa-after-creation is a GATEKEEPER task:
+    - Primary purpose: Pass/Fail decision after component creation
+    - Secondary purpose: Quality score for tracking
+
+    For pass/fail: HAIKU MATCHES OPUS ✓
+    For score precision: HAIKU INFLATES (+19%)
+
+    Recommendation: Use Haiku for QA gating decisions.
+    If precise score needed, use Opus or add calibration.
+
+  learning:
+    pattern_identified: |
+      GATEKEEPER tasks (binary pass/fail) are HAIKU-ELIGIBLE
+      even if scores differ, as long as final decision matches.
+
+    similar_tasks:
+      - "qa-after-creation (qualified)"
+      - "validate-extraction (likely qualified)"
+      - "security-scan (likely qualified)"
+
+    different_from:
+      - "validate-squad (requires type detection)"
+      - "pv-modernization-score (requires interpretation)"
+
+metadata:
+  validator: "@pedro-valerio"
+  note: "Output files not created by agents (I/O issue). Report based on completion notifications."
+  tokens_opus: 113793
+  tokens_haiku: 59289
+  total_cost: "$0.69"

package/pro/squads/squad-creator-pro/test-cases/smoke-test-model-routing/test-case.yaml

@@ -1,100 +1,100 @@
-# Test Case: smoke-test-model-routing
-# Model Tier Qualification - Wave 3A
-# Similar to: validate-squad (hybrid with worker scripts)
-
-test_case:
-  name: "smoke-test-model-routing-standard"
-  version: "1.0"
-  created: "2026-02-11"
-  task_file: "tasks/smoke-test-model-routing.md"
-  wave: "3A"
-  priority: 2
-  similar_to: "validate-squad"
-  confidence: "HIGH"
-
-input:
-  # Test model routing for a known task
-  target: "squads/squad-creator-pro/tasks/pv-axioma-assessment.md"
-  params:
-    expected_tier: "haiku"
-  context_files:
-    - "squads/squad-creator-pro/config/model-routing.yaml"
-
-expected_output:
-  sections:
-    - routing_result
-    - validation_checks
-    - tier_decision
-    - evidence
-
-  fields:
-    - name: "detected_tier"
-      type: "enum"
-      values: ["haiku", "sonnet", "opus", "script"]
-    - name: "matches_expected"
-      type: "boolean"
-    - name: "validation_passed"
-      type: "boolean"
-
-  validations:
-    - "Deve detectar tier baseado em task metadata"
-    - "Deve validar contra model-routing.yaml"
-    - "Deve reportar match/mismatch"
-    - "Scripts devem rodar antes do LLM"
-
-baseline:
-  opus_output: "opus-baseline.yaml"
-  opus_score: 10.0
-  opus_tokens:
-    input: 0
-    output: 0
-  opus_latency_ms: 0
-
-evaluation:
-  completeness:
-    required_sections:
-      - "detected_tier: resultado da detecção"
-      - "validation_checks: lista de checks"
-      - "evidence: justificativa"
-    optional_sections:
-      - "recommendations: se mismatch"
-
-  accuracy:
-    numeric_tolerance: 0.05
-    enum_exact_match: true
-
-  reasoning:
-    minimum_justifications: 3
-    requires_evidence: true
-
-  format:
-    expected_format: "yaml"
-    parseable: true
-
-  actionability:
-    minimum_recommendations: 0
-    specificity: "high"
-
-compensation_history:
-  attempts: []
-
-results:
-  haiku:
-    score: null
-    percentage: null
-    qualified: null
-    output_file: "haiku-output.yaml"
-
-  opus:
-    score: 10.0
-    percentage: 100
-    qualified: true
-    output_file: "opus-baseline.yaml"
-
-  final_tier: null
-  recommendation: null
-
-wave_3a_notes:
-  rationale: "Hybrid task with worker scripts, similar to validate-squad (91% match)"
-  expected_fixes: "None - scripts do heavy lifting"
-  estimated_effort: "20 min"
+# Test Case: smoke-test-model-routing
+# Model Tier Qualification - Wave 3A
+# Similar to: validate-squad (hybrid with worker scripts)
+
+test_case:
+  name: "smoke-test-model-routing-standard"
+  version: "1.0"
+  created: "2026-02-11"
+  task_file: "tasks/smoke-test-model-routing.md"
+  wave: "3A"
+  priority: 2
+  similar_to: "validate-squad"
+  confidence: "HIGH"
+
+input:
+  # Test model routing for a known task
+  target: "squads/squad-creator-pro/tasks/pv-axioma-assessment.md"
+  params:
+    expected_tier: "haiku"
+  context_files:
+    - "squads/squad-creator-pro/config/model-routing.yaml"
+
+expected_output:
+  sections:
+    - routing_result
+    - validation_checks
+    - tier_decision
+    - evidence
+
+  fields:
+    - name: "detected_tier"
+      type: "enum"
+      values: ["haiku", "sonnet", "opus", "script"]
+    - name: "matches_expected"
+      type: "boolean"
+    - name: "validation_passed"
+      type: "boolean"
+
+  validations:
+    - "Deve detectar tier baseado em task metadata"
+    - "Deve validar contra model-routing.yaml"
+    - "Deve reportar match/mismatch"
+    - "Scripts devem rodar antes do LLM"
+
+baseline:
+  opus_output: "opus-baseline.yaml"
+  opus_score: 10.0
+  opus_tokens:
+    input: 0
+    output: 0
+  opus_latency_ms: 0
+
+evaluation:
+  completeness:
+    required_sections:
+      - "detected_tier: resultado da detecção"
+      - "validation_checks: lista de checks"
+      - "evidence: justificativa"
+    optional_sections:
+      - "recommendations: se mismatch"
+
+  accuracy:
+    numeric_tolerance: 0.05
+    enum_exact_match: true
+
+  reasoning:
+    minimum_justifications: 3
+    requires_evidence: true
+
+  format:
+    expected_format: "yaml"
+    parseable: true
+
+  actionability:
+    minimum_recommendations: 0
+    specificity: "high"
+
+compensation_history:
+  attempts: []
+
+results:
+  haiku:
+    score: null
+    percentage: null
+    qualified: null
+    output_file: "haiku-output.yaml"
+
+  opus:
+    score: 10.0
+    percentage: 100
+    qualified: true
+    output_file: "opus-baseline.yaml"
+
+  final_tier: null
+  recommendation: null
+
+wave_3a_notes:
+  rationale: "Hybrid task with worker scripts, similar to validate-squad (91% match)"
+  expected_fixes: "None - scripts do heavy lifting"
+  estimated_effort: "20 min"

package/pro/squads/squad-creator-pro/test-cases/upgrade-squad/test-case.yaml

@@ -1,106 +1,106 @@
-# Test Case: upgrade-squad
-# Model Tier Qualification - Wave 3A
-# Similar to: validate-squad (uses 3 checklists, script for inventory)
-
-test_case:
-  name: "upgrade-squad-standard"
-  version: "1.0"
-  created: "2026-02-11"
-  task_file: "tasks/upgrade-squad.md"
-  wave: "3A"
-  priority: 3
-  similar_to: "validate-squad"
-  confidence: "HIGH"
-
-input:
-  # Use a real squad to analyze for upgrades
-  target: "squads/squad-creator"
-  params:
-    mode: "audit"  # audit only, no changes
-    focus: "agents"
-  context_files:
-    - "squads/squad-creator-pro/checklists/agent-quality-gate.md"
-    - "squads/squad-creator-pro/checklists/task-anatomy-checklist.md"
-    - "squads/squad-creator-pro/checklists/squad-checklist.md"
-
-expected_output:
-  sections:
-    - inventory
-    - gap_analysis
-    - upgrade_plan
-    - priority_list
-
-  fields:
-    - name: "components_analyzed"
-      type: "number"
-      range: [1, 200]
-    - name: "components_passing"
-      type: "number"
-    - name: "components_needing_upgrade"
-      type: "number"
-    - name: "overall_status"
-      type: "enum"
-      values: ["PASS", "NEEDS_UPGRADE", "CRITICAL"]
-
-  validations:
-    - "Deve inventariar todos os componentes"
-    - "Deve aplicar checklists específicos por tipo"
-    - "Deve classificar gaps por prioridade (critical/high/medium/low)"
-    - "Deve estimar esforço por componente"
-
-baseline:
-  opus_output: "opus-baseline.yaml"
-  opus_score: 10.0
-  opus_tokens:
-    input: 0
-    output: 0
-  opus_latency_ms: 0
-
-evaluation:
-  completeness:
-    required_sections:
-      - "inventory: contagem e listagem de componentes"
-      - "gap_analysis: gaps por componente"
-      - "priority_list: ordenado por criticidade"
-    optional_sections:
-      - "upgrade_plan: plano de execução"
-
-  accuracy:
-    numeric_tolerance: 0.10
-    enum_exact_match: true
-
-  reasoning:
-    minimum_justifications: 5
-    requires_evidence: true
-
-  format:
-    expected_format: "yaml"
-    parseable: true
-
-  actionability:
-    minimum_recommendations: 3
-    specificity: "high"
-
-compensation_history:
-  attempts: []
-
-results:
-  haiku:
-    score: null
-    percentage: null
-    qualified: null
-    output_file: "haiku-output.yaml"
-
-  opus:
-    score: 10.0
-    percentage: 100
-    qualified: true
-    output_file: "opus-baseline.yaml"
-
-  final_tier: null
-  recommendation: null
-
-wave_3a_notes:
-  rationale: "Uses 3 existing checklists, script for inventory - checklist-based"
-  expected_fixes: "Scope definition for agent analysis"
-  estimated_effort: "45 min"
+# Test Case: upgrade-squad
+# Model Tier Qualification - Wave 3A
+# Similar to: validate-squad (uses 3 checklists, script for inventory)
+
+test_case:
+  name: "upgrade-squad-standard"
+  version: "1.0"
+  created: "2026-02-11"
+  task_file: "tasks/upgrade-squad.md"
+  wave: "3A"
+  priority: 3
+  similar_to: "validate-squad"
+  confidence: "HIGH"
+
+input:
+  # Use a real squad to analyze for upgrades
+  target: "squads/squad-creator"
+  params:
+    mode: "audit"  # audit only, no changes
+    focus: "agents"
+  context_files:
+    - "squads/squad-creator-pro/checklists/agent-quality-gate.md"
+    - "squads/squad-creator-pro/checklists/task-anatomy-checklist.md"
+    - "squads/squad-creator-pro/checklists/squad-checklist.md"
+
+expected_output:
+  sections:
+    - inventory
+    - gap_analysis
+    - upgrade_plan
+    - priority_list
+
+  fields:
+    - name: "components_analyzed"
+      type: "number"
+      range: [1, 200]
+    - name: "components_passing"
+      type: "number"
+    - name: "components_needing_upgrade"
+      type: "number"
+    - name: "overall_status"
+      type: "enum"
+      values: ["PASS", "NEEDS_UPGRADE", "CRITICAL"]
+
+  validations:
+    - "Deve inventariar todos os componentes"
+    - "Deve aplicar checklists específicos por tipo"
+    - "Deve classificar gaps por prioridade (critical/high/medium/low)"
+    - "Deve estimar esforço por componente"
+
+baseline:
+  opus_output: "opus-baseline.yaml"
+  opus_score: 10.0
+  opus_tokens:
+    input: 0
+    output: 0
+  opus_latency_ms: 0
+
+evaluation:
+  completeness:
+    required_sections:
+      - "inventory: contagem e listagem de componentes"
+      - "gap_analysis: gaps por componente"
+      - "priority_list: ordenado por criticidade"
+    optional_sections:
+      - "upgrade_plan: plano de execução"
+
+  accuracy:
+    numeric_tolerance: 0.10
+    enum_exact_match: true
+
+  reasoning:
+    minimum_justifications: 5
+    requires_evidence: true
+
+  format:
+    expected_format: "yaml"
+    parseable: true
+
+  actionability:
+    minimum_recommendations: 3
+    specificity: "high"
+
+compensation_history:
+  attempts: []
+
+results:
+  haiku:
+    score: null
+    percentage: null
+    qualified: null
+    output_file: "haiku-output.yaml"
+
+  opus:
+    score: 10.0
+    percentage: 100
+    qualified: true
+    output_file: "opus-baseline.yaml"
+
+  final_tier: null
+  recommendation: null
+
+wave_3a_notes:
+  rationale: "Uses 3 existing checklists, script for inventory - checklist-based"
+  expected_fixes: "Scope definition for agent analysis"
+  estimated_effort: "45 min"