aiox-core 5.0.2 → 5.0.4

package/pro/squads/squad-creator-pro/test-cases/an-assess-sources/qualification-report.yaml

@@ -0,0 +1,213 @@
+# Qualification Report: an-assess-sources v2.0
+# Task: Model Tier Qualification Test
+# Date: 2026-02-11
+
+qualification_test:
+  task: "an-assess-sources"
+  task_version: "2.0.0"
+  test_date: "2026-02-11"
+  target: "outputs/minds/alex_hormozi"
+
+# ============================================================================
+# TEST RESULTS
+# ============================================================================
+
+results:
+  opus_baseline:
+    model: "opus"
+    assessment_level: "BOOK (whole books)"
+    sources_assessed: 4
+    tier_distribution:
+      crown_jewel: 3  # All 3 books
+      ouro: 0
+      mixed: 0
+      bronze: 1       # Tom Bilyeu (no content)
+    average_quality: 4.15
+    scores:
+      $100M_Offers: { score: 4.8, tier: "crown_jewel" }
+      $100M_Leads: { score: 5.0, tier: "crown_jewel" }
+      $100M_Money_Models: { score: 5.0, tier: "crown_jewel" }
+      Tom_Bilyeu: { score: 0.0, tier: "bronze" }
+
+  haiku_test:
+    model: "haiku"
+    assessment_level: "CHAPTER (granular)"
+    sources_assessed: 5  # Treated chapters as separate sources
+    tier_distribution:
+      crown_jewel: 1  # Only Cap 6
+      ouro: 3         # Cap 8, Leads I-II, Money Models
+      mixed: 0
+      bronze: 1       # Tom Bilyeu (no content)
+    average_quality: 4.2
+    scores:
+      Cap_6_Equacao_Valor: { score: 5.0, tier: "crown_jewel" }
+      Cap_8_Pensamento: { score: 4.8, tier: "ouro" }
+      Leads_Secoes_I_II: { score: 4.8, tier: "ouro" }
+      Money_Models: { score: 4.4, tier: "ouro" }
+      Tom_Bilyeu: { score: 0.2, tier: "bronze" }
+
+# ============================================================================
+# COMPARISON ANALYSIS
+# ============================================================================
+
+comparison:
+  # Scope Interpretation Issue (Primary)
+  scope_interpretation:
+    issue: "CRITICAL - Models interpreted scope differently"
+    opus_interpretation: "Assess at BOOK level (whole book = 1 source)"
+    haiku_interpretation: "Assess at CHAPTER level (chapters = separate sources)"
+    impact: "Cannot directly compare scores - different units of analysis"
+    root_cause: "Task v2.0 did not explicitly define assessment granularity"
+
+  # Tier Distribution
+  tier_match:
+    tom_bilyeu: "MATCH - Both BRONZE (no transcription)"
+    books_overall: "MISMATCH"
+    opus_books: "3 CROWN JEWEL"
+    haiku_books: "1 CROWN JEWEL + 3 OURO"
+
+  # Score Variance (where comparable)
+  score_comparison:
+    note: "Scores not directly comparable due to granularity mismatch"
+    tom_bilyeu:
+      opus: 0.0
+      haiku: 0.2
+      variance: "+0.2 (Haiku slightly generous - gave 1 PASS)"
+    average_quality:
+      opus: 4.15
+      haiku: 4.2
+      variance: "+1.2% (within tolerance)"
+
+  # Binary Checkpoint Effectiveness
+  binary_checkpoints:
+    observation: "Both used 25 checkpoints correctly"
+    scoring_philosophy_applied: true
+    checkpoint_documentation: "Both provided detailed notes per checkpoint"
+    variance_source: "Not in scoring - in scope interpretation"
+
+# ============================================================================
+# VETO CONDITION CHECK
+# ============================================================================
+
+veto_conditions:
+  MTQ_VC_001:  # Score differs by >10%
+    triggered: false
+    analysis: "Average quality 4.15 vs 4.2 = 1.2% variance"
+
+  MTQ_VC_004:  # Different tier classification
+    triggered: true
+    analysis: |
+      - Opus: 3 books → all CROWN JEWEL
+      - Haiku: Same content → 1 CROWN JEWEL + 3 OURO
+      - Root cause: Scope interpretation (book vs chapter level)
+
+# ============================================================================
+# DECISION
+# ============================================================================
+
+decision:
+  verdict: "OPUS REQUIRED"
+  confidence: "high"
+
+  rationale: |
+    1. SCOPE INTERPRETATION ISSUE
+       - Opus assessed at book level (aggregated)
+       - Haiku assessed at chapter level (granular)
+       - Task v2.0 did not specify assessment granularity
+       - This is similar to an-clone-review scope issue (wrapper vs delegated persona)
+
+    2. TIER CLASSIFICATION MISMATCH
+       - 3 Crown Jewel (Opus) vs 1 Crown Jewel (Haiku)
+       - Veto condition MTQ_VC_004 triggered
+       - Different tier = different extraction priorities for user
+
+    3. BINARY CHECKPOINTS WORKED FOR SCORING
+       - Both models scored individual checkpoints consistently
+       - The 25 binary checkpoints eliminated subjective scoring variance
+       - Variance came from scope interpretation, not scoring calibration
+
+    4. PATTERN MATCH
+       - Similar to an-clone-review (Opus contextual, Haiku literal)
+       - Haiku followed instructions literally (more granular)
+       - Opus inferred user intent (book-level assessment)
+
+  required_fix: |
+    Update task v2.1 with explicit scope definition:
+
+    ## SCOPE DEFINITION (CRITICAL)
+
+    ```yaml
+    assessment_granularity:
+      books: "Assess ENTIRE BOOK as single source (not chapters)"
+      videos: "Assess ENTIRE VIDEO as single source (not segments)"
+      podcasts: "Assess ENTIRE EPISODE as single source"
+      posts: "Each post is one source"
+
+      rationale: |
+        Extraction priority is per SOURCE, not per chapter.
+        User downloads/processes at source level.
+        Crown Jewel at book level = prioritize that book.
+    ```
+
+# ============================================================================
+# PATTERN DISCOVERY
+# ============================================================================
+
+pattern_discovered:
+  name: "Granularity Interpretation"
+  description: |
+    When task doesn't specify assessment granularity:
+    - Opus infers user intent (book-level for books)
+    - Haiku follows instructions literally (more granular)
+
+  fix_pattern: |
+    Add explicit granularity rules to tasks:
+    - "Assess at {LEVEL} granularity"
+    - "One source = one {UNIT}"
+    - "Do NOT break into sub-units"
+
+  affected_tasks:
+    - an-assess-sources (source granularity)
+    - an-clone-review (scope of review)
+    - validate-squad (type detection scope)
+
+# ============================================================================
+# METRICS
+# ============================================================================
+
+metrics:
+  test_coverage:
+    sources_tested: 4
+    dimensions_covered: 5
+    checkpoints_evaluated: "25 × 4 = 100 total"
+
+  execution:
+    opus_time: "~60s"
+    haiku_time: "~25s"
+
+  binary_checkpoint_effectiveness:
+    score_variance: "1.2% (excellent)"
+    tier_match_rate: "25% (1 of 4 sources - Tom Bilyeu)"
+
+  root_cause:
+    scoring: "0% of variance"
+    scope_interpretation: "100% of variance"
+
+# ============================================================================
+# NEXT STEPS
+# ============================================================================
+
+next_steps:
+  - action: "Update task to v2.1 with scope definition"
+    priority: "high"
+
+  - action: "Add to model-routing.yaml: an-assess-sources → opus"
+    priority: "high"
+
+  - action: "Update BATCH-PROGRESS.md with result"
+    priority: "medium"
+
+  - action: "Consider if scope fix could enable Haiku (future test)"
+    priority: "low"
+    note: "Would require re-testing with v2.1"
+

package/pro/squads/squad-creator-pro/test-cases/an-assess-sources/test-case.yaml

@@ -0,0 +1,69 @@
+# Test Case: an-assess-sources v2.0
+# Target: Alex Hormozi mind
+# Purpose: Validate Binary Checkpoint Conversion for Haiku compatibility
+
+test_case:
+  task: "an-assess-sources"
+  task_version: "2.0.0"
+  test_date: "2026-02-11"
+  workflow_version: "wf-model-tier-qualification v2.0"
+
+target:
+  mind: "alex_hormozi"
+  mind_path: "outputs/minds/alex_hormozi"
+
+# Sources to assess (subset for test - representative sample)
+sources_for_assessment:
+  - name: "Livro $100M Offers"
+    type: "livro"
+    path: "outputs/minds/alex_hormozi/sources/01 Livros/Livro 100M Ofertas/"
+    description: "18 capítulos - Framework central de criação de ofertas"
+    estimated_duration: "~200 páginas"
+
+  - name: "Livro $100M Leads"
+    type: "livro"
+    path: "outputs/minds/alex_hormozi/sources/01 Livros/Livro 100M Leads/"
+    description: "5 seções - Geração de leads e aquisição"
+    estimated_duration: "~250 páginas"
+
+  - name: "Livro $100M Money Models"
+    type: "livro"
+    path: "outputs/minds/alex_hormozi/sources/01 Livros/Livro 100m Modelos/"
+    description: "8 seções - Modelos de negócio e monetização"
+    estimated_duration: "~150 páginas"
+
+  - name: "Entrevista Tom Bilyeu (Impact Theory)"
+    type: "entrevista"
+    path: "outputs/minds/alex_hormozi/sources/interviews/Entrevista Tom Biley.md"
+    description: "Entrevista longa com perguntas profundas"
+    estimated_duration: "~2 horas"
+
+# Qualification criteria
+qualification:
+  veto_conditions:
+    - "MTQ_VC_001: Haiku score differs by >10% from Opus"
+    - "MTQ_VC_004: Different tier classification"
+
+  expected_outcome:
+    with_binary_checkpoints: "High match rate (>95%)"
+    rationale: "25 binary checkpoints should eliminate subjective scoring variance"
+
+# Test execution plan
+execution:
+  phase_1_opus_baseline:
+    model: "opus"
+    output: "opus-baseline.yaml"
+    purpose: "Establish reference scores"
+
+  phase_2_haiku_test:
+    model: "haiku"
+    output: "haiku-output.yaml"
+    purpose: "Test Haiku compatibility with v2.0 binary checkpoints"
+
+  phase_3_comparison:
+    output: "qualification-report.yaml"
+    metrics:
+      - "Per-source tier match"
+      - "Per-dimension score variance"
+      - "Overall classification match"
+      - "Crown Jewel identification match"

package/pro/squads/squad-creator-pro/test-cases/an-clone-review/haiku-round-1.yaml

@@ -0,0 +1,213 @@
+clone_review:
+  clone: mmos-victoria
+  clone_file: .claude/agents/mmos-victoria.md
+  date: "2026-02-11"
+
+  source_quality:
+    total_sources: 2
+    ouro_count: 1
+    bronze_count: 1
+    ouro_percentage: "50%"
+    curadoria_score: "aceitavel"
+    sources_checked:
+      - file: mmos-victoria.md
+        tier: bronze
+        checkpoints_passed: [1, 2, 3, 4, 5]
+        notes: "Minimal agent wrapper (38 lines), delegates to persona file"
+      - file: victoria-viability-specialist.md
+        tier: ouro
+        checkpoints_passed: [1, 2, 3, 4, 5]
+        notes: "Comprehensive persona (700 lines), complete DNA, frameworks, and patterns"
+
+  trinity:
+    playbook:
+      checkpoints_passed: [1, 2, 3, 4, 5]
+      checkpoints_failed: []
+      score: 5
+      evidence:
+        "1_section_exists": "STRICT RULES section with Step 1, 2, 3"
+        "2_has_steps": "3 sequential steps: Load Context, Load Persona, Execute Mission"
+        "3_sequential": "Numbered ## Step 1/2/3 with clear progression"
+        "4_actionable": "Action verbs: Load, Parse, Read, Execute, Adopt"
+        "5_output": "Parse outputs, adopt persona, execute from prompt"
+      gaps: []
+    framework:
+      checkpoints_passed: [1, 2, 3, 4, 5]
+      checkpoints_failed: []
+      score: 5
+      evidence:
+        "1_section_exists": "STRICT RULES section with constraints"
+        "2_has_rules": "5 rules: FIRST tool call, SECOND tool call, NEVER skip, context loading, Parse output"
+        "3_conditional": "MUST conditions on first/second call order"
+        "4_actions": "MUST be, NEVER skip, Parse JSON output"
+        "5_tree": "If context missing → error shown; If command parsed → execute with criteria"
+      gaps: []
+    swipe_file:
+      checkpoints_passed: [5]
+      checkpoints_failed: [1, 2, 3, 4]
+      score: 1
+      evidence:
+        "1_section_missing": "No examples section in agent file (delegated to persona)"
+        "2_examples_missing": "No concrete examples in agent wrapper"
+        "3_citations_missing": "No SOURCE tags"
+        "4_pairs_missing": "No input/output pairs shown"
+        "5_diverse": "References multiple file types (.md, .yaml, .cjs)"
+      gaps:
+        - "Agent wrapper has no example interactions"
+        - "Missing before/after examples of Load/Parse/Execute"
+        - "Persona file HAS examples, but wrapper doesn't reference them"
+        - "No workflow trace showing actual context loading output"
+    total_score: 11
+    max_score: 15
+    percentage: "73.3%"
+    verdict: "NEEDS_WORK"
+    analysis: |
+      Trinity is strong in Playbook and Framework (both 5/5), but severely lacking
+      in Swipe File. The minimal agent wrapper delegates all persona details to a
+      separate 700-line persona file. This is ARCHITECTURALLY CORRECT but means the
+      Swipe File examples exist only in the persona, not in the agent itself.
+
+      Gap: The agent wrapper needs to show HOW the persona is invoked with examples,
+      or at minimum reference where examples are found (victoria-viability-specialist.md).
+
+  stages:
+    has_stages: false
+    needs_stages: true
+    stage_analysis: |
+      Agent uses implicit 3-stage workflow (Load → Load → Execute) but doesn't
+      formalize as explicit "stages:" section. Given 9 context references and
+      variable behavior based on context state, this architecture would benefit
+      from explicit stage definitions.
+    stage_score: 0
+    gaps:
+      - "No explicit 'stages:' section despite 3-step workflow"
+      - "Stage transitions (Load Context → Load Persona → Execute) are implicit, not formal"
+      - "No trigger conditions on stages (FIRST/SECOND call rules are conditions, not triggers)"
+      - "Missing explicit per-stage behavior blocks"
+
+  quick_fidelity:
+    checkpoints_passed: [1, 2, 3, 4, 5]
+    checkpoints_failed: []
+    score: 5
+    percentage: "100%"
+    evidence:
+      "1_voice_dna": "Victoria persona has sentence_starters, metaphors, vocabulary"
+      "2_phrases": "3+ signature methods: Jobs-to-be-Done, Deployment Architecture, Integration Testing"
+      "3_thinking_dna": "Core principles: christensen_foundation, viability_methodology, deployment_philosophy"
+      "4_frameworks": "4+ frameworks documented: JTBD, Christensen, Use Case, Deployment, Testing, Handoff"
+      "5_immune": "Anti-patterns section with 10 clear veto conditions (AVOID: ...)"
+    estimate: "Advanced (V3.0+)"
+    notes: |
+      The persona file (victoria-viability-specialist.md) is exceptionally comprehensive.
+      This is NOT a thin wrapper - it's a 700-line complete agent definition with:
+      - 5 complete signature methods (each 20-40 lines)
+      - 3 behavioral states with triggers
+      - 10 objection algorithms with detailed responses
+      - 3 output examples with full input/output
+      - 8 validation gates
+      - Complete quality standards and methodological toolkit
+
+  overall:
+    trinity_percentage: "73.3%"
+    fidelity_percentage: "100%"
+    verdict: "NEEDS_WORK"
+    composite_score: "86.7%"
+    explanation: |
+      Paradox: Agent file is tiny (38 lines) but delegates to a MASSIVE persona file
+      (700 lines). If we evaluate just the agent wrapper, it's 73% (NEEDS_WORK due to
+      missing Swipe examples). But if we evaluate the SYSTEM (agent + persona), it's
+      exceptional - the persona has advanced fidelity, complete patterns, and 100 examples.
+
+      RECOMMENDATION: This is a TWO-PART AGENT system where separation of concerns is
+      intentional. The agent wrapper is correct for its role (context loader + persona
+      adopter). The persona is production-ready. But the agent wrapper should document
+      that it's Part 1 of 2 and show at least one example of the full workflow.
+
+  priority_actions:
+    - action: "Add Swipe File examples to mmos-victoria.md wrapper"
+      target_component: swipe_file
+      checkpoints_to_fix: [1, 2, 3, 4]
+      impact: "alto"
+      description: "Show actual workflow of context loading and persona adoption with real input/output pairs"
+      example: |
+        ## Example Workflow
+
+        **Input:** User spawns mmos-victoria
+        **Output:** Context loader reads .active-mind (ztest15_dbg), returns JSON with slug, person_name, mode
+        **Victoria Response:** Adopts persona, greets user, awaits command
+
+    - action: "Formalize stages with explicit stage definitions"
+      target_component: stages
+      checkpoints_to_fix: [1, 2, 3, 4, 5]
+      impact: "medio"
+      description: "Convert implicit 3-step workflow to formal 'stages:' section with triggers and behaviors"
+      rationale: "Improves clarity and enables other systems to introspect stage flow"
+
+    - action: "Cross-reference persona examples in agent wrapper"
+      target_component: swipe_file
+      checkpoints_to_fix: [1, 2]
+      impact: "medio"
+      description: "Add 'See also: victoria-viability-specialist.md for complete examples' with line ranges"
+
+  next_version_path: |
+    Current: 73% Trinity + 100% Fidelity = 86.7% composite
+    Target: 90%+ Trinity (close Swipe File gap) + 100% Fidelity = 95% composite
+
+    To reach 90% Trinity (13/15):
+    - Add 3+ examples to Swipe File (checkpoint 2) → +1 point
+    - Add SOURCE citations for examples (checkpoint 3) → +1 point
+    - Add input/output pairs (checkpoint 4) → +1 point
+
+    Effort: ~30 minutes. Add 15-20 lines to mmos-victoria.md showing workflow examples.
+
+metadata:
+  task_version: "2.0.0"
+  scoring_method: "Binary checkpoint counting"
+  model_compatible: ["opus", "sonnet", "haiku"]
+  execution_date: "2026-02-11"
+  reviewer_agent: "haiku"
+
+  checkpoint_distribution:
+    source_quality:
+      checkpoints: 5
+      passed: 4
+      percentage: "80%"
+    trinity_playbook:
+      checkpoints: 5
+      passed: 5
+      percentage: "100%"
+    trinity_framework:
+      checkpoints: 5
+      passed: 5
+      percentage: "100%"
+    trinity_swipe:
+      checkpoints: 5
+      passed: 1
+      percentage: "20%"
+    fidelity:
+      checkpoints: 5
+      passed: 5
+      percentage: "100%"
+
+  validation_summary:
+    passes: "Trinity playbook/framework are bulletproof. Fidelity is exceptional."
+    issues: "Swipe File missing from agent wrapper (exists in persona only). Stages not formalized."
+    confidence: "HIGH - All checkpoints verified by grep/line counting"
+
+  findings:
+    architectural_pattern: "Two-part agent system: minimal wrapper + comprehensive persona"
+    strengths:
+      - "Perfect playbook implementation (3-step workflow)"
+      - "Strict rules enforce correct execution order (FIRST/SECOND call)"
+      - "Exceptional fidelity in persona (advanced V3.0)"
+      - "Complete anti-pattern documentation"
+    weaknesses:
+      - "Swipe File examples only in persona, not agent wrapper"
+      - "Stages implicit, not explicit"
+      - "No example showing Load Context output"
+      - "No trace of actual workflow execution"
+
+  files_involved:
+    - ".claude/agents/mmos-victoria.md (38 lines)"
+    - ".claude/agents/MMOS/agents/victoria-viability-specialist.md (700 lines)"
+    - "squads/mmos/scripts/mmos-context-loader.cjs (referenced)"