Category C: Design Agents

Paradigm-specific design consultation and meta-analysis orchestration with multi-gate validation — 4 agents

Paradigm Coverage

Paradigm-specific (C1 Quantitative, C2 Qualitative, C3 Mixed Methods), Meta-analysis (C5 — now includes effect size extraction, data integrity, and error prevention)

C5 Meta-Analysis Master (v11.0)

Based on V7 GenAI meta-analysis lessons learned: Consolidated single-agent architecture with decision authority, data integrity validation, and error prevention (formerly C5/C6/C7).

Meta-Analysis Architecture
┌─────────────────────────────────────────────────┐
│ C5-MetaAnalysisMaster (Consolidated v11.0) │
│ └─ Multi-gate validation (4 gates) │
│ └─ Phase-based orchestration (7 phases) │
│ └─ ES hierarchy enforcement │
│ └─ Effect size extraction (ex-B3) │
│ └─ Hedges' g calc + SD recovery (ex-C6) │
│ └─ Error prevention + anomaly detect (ex-C7) │
│ └─ Sensitivity analysis │
└─────────────────────────────────────────────────┘

Multi-Gate Validation (C5)

Four-gate validation workflow ensures data integrity:

1

Extraction Validation

Verify data completeness and accuracy

2

Classification Validation

ES Hierarchy enforcement (pre-test exclusion)

3

Statistical Validation

Hedges' g calculation with SD recovery

4

Independence Validation

Ensure statistical independence, exclude pre-tests

C1 - Quantitative Design & Sampling

Creative quantitative design options with sampling strategies and experimental materials development

Tier:HIGH
Model:Opus
VS Mode:Enhanced VS 3-Phase

Capabilities

  • ▸Enhanced VS 3-Phase methodology
  • ▸RCT and quasi-experimental designs
  • ▸Survey design (cross-sectional, longitudinal)
  • ▸Factorial designs (between, within, mixed)
  • ▸Power analysis and sample size calculation
  • ▸Randomization and sampling strategies
  • ▸Treatment protocol and control condition design (formerly C4)
  • ▸Manipulation check design and fidelity monitoring (formerly C4)

Auto-Trigger Keywords

English"experimental design", "RCT", "quasi-experimental", "survey design", "power analysis", "experimental materials", "treatment design"
한국어"실험 설계", "RCT", "준실험", "조사 설계", "검정력 분석", "실험 자료", "처치 설계"

Human Checkpoints

🔴 CP_METHODOLOGY_APPROVAL

C2 - Qualitative Design

Comprehensive qualitative research design across major traditions, including ethnography and action research

Tier:HIGH
Model:Opus
VS Mode:Enhanced VS 3-Phase

Capabilities

  • ▸Enhanced VS 3-Phase methodology
  • ▸Phenomenology (descriptive, hermeneutic)
  • ▸Grounded theory (Straussian, Glaserian, Constructivist)
  • ▸Case study design (single, multiple, embedded)
  • ▸Narrative inquiry
  • ▸Ethnography and participatory action research
  • ▸Sample size justification for qualitative

Auto-Trigger Keywords

English"qualitative design", "phenomenology", "grounded theory", "case study", "ethnography", "action research"
한국어"질적 연구 설계", "현상학", "근거이론", "사례연구", "문화기술지", "실행연구"

Human Checkpoints

🔴 CP_METHODOLOGY_APPROVAL

C3 - Mixed Methods Design Consultant

Sequential, concurrent, and embedded mixed methods designs

Tier:HIGH
Model:Opus
VS Mode:Standard

Capabilities

  • ▸Sequential designs (explanatory, exploratory)
  • ▸Convergent parallel designs
  • ▸Embedded designs
  • ▸Transformative frameworks
  • ▸Morse notation support (QUAL→quan, QUAN+qual)
  • ▸Integration timing and weighting decisions

Auto-Trigger Keywords

English"mixed methods", "sequential design", "convergent", "QUAL", "QUAN"
한국어"혼합방법", "혼합 연구", "순차적 설계", "통합"

Human Checkpoints

🔴 CP_METHODOLOGY_APPROVAL

C5 - Meta-Analysis Master

Decision Authority

Multi-gate validation workflow orchestration with effect size extraction, data integrity, error prevention, and sensitivity analysis

Tier:HIGH
Model:Opus
VS Mode:Full VS 5-Phase

Capabilities

  • ▸Multi-gate validation (4 gates)
  • ▸Phase-based orchestration (7 phases)
  • ▸ES hierarchy enforcement
  • ▸Pre-test exclusion protocol
  • ▸Hedges' g pooling and effect size extraction (formerly B3)
  • ▸Heterogeneity analysis (I², τ², Q-statistic)
  • ▸Publication bias assessment
  • ▸Data integrity validation and Hedges' g calculation with SD recovery (formerly C6)
  • ▸Error prevention, anomaly detection, and pre-test pattern detection (formerly C7)
  • ▸Sensitivity analysis and robustness checks

Auto-Trigger Keywords

English"meta-analysis", "pooled effect", "heterogeneity", "effect size synthesis", "effect size", "data extraction", "error prevention"
한국어"메타분석", "메타 분석", "통합 효과", "이질성", "효과크기", "데이터 추출", "오류 방지"

Human Checkpoints

🔴 CP_META_GATE (any gate failure)🔴 META_TIER3_REVIEW (data completeness < 40%)🟠 META_ANOMALY_REVIEW (|g| > 3.0)🟠 META_PRETEST_CONFIRM (ambiguous classification)

Category C agents (4 agents) use Enhanced VS 3-Phase methodology for paradigm-specific designs (C1/C2/C3) and Full VS 5-Phase for meta-analysis (C5). C5 now consolidates the former C5/C6/C7 three-tier architecture into a single agent with decision authority, data integrity validation, and error prevention. Experimental materials (formerly C4) are now part of C1.