Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Diverga help guide - displays all 24 agents across 9 categories, commands, and usage examples. Triggers: help, guide, how to use, 도움말
.claude/skills/brycewang-stanford-help/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 10 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-15 | ✗→✓ | ▲ Improved | 157% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 55% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 175% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 121% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 44% | 0% |
Version: 2.0.0 Trigger: /diverga:help
Displays comprehensive guide for Diverga, including all 24 agents across 9 categories, commands, and usage examples.
When user invokes /diverga:help, display:
╔══════════════════════════════════════════════════════════════════╗
║ Diverga v11.0 Help ║
║ AI Research Assistant - 24 Agents, 9 Categories ║
╚══════════════════════════════════════════════════════════════════╝
┌─────────────────────────────────────────────────────────────────┐
│ QUICK START │
├─────────────────────────────────────────────────────────────────┤
│ Just describe your research: │
│ "I want to conduct a meta-analysis on AI in education" │
│ "Help me design a qualitative study" │
│ "메타분석 연구를 시작하고 싶어" │
│ │
│ Diverga auto-detects context and activates relevant agents. │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ COMMANDS │
├─────────────────────────────────────────────────────────────────┤
│ /diverga:setup Initial configuration wizard │
│ /diverga:doctor System diagnostics & health check │
│ /diverga:help This help guide │
│ /diverga:meta-analysis Meta-analysis workflow (C5) │
│ /diverga:humanize Humanization pipeline (G5+G6+F5) │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY A: FOUNDATION (3 agents) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:a1 ResearchQuestionRefiner Refine research Qs │
│ diverga:a2 TheoreticalFrameworkArchitect Frameworks + Critique │
│ + Visualization (absorbed A3, A6) │
│ diverga:a5 ParadigmWorldviewAdvisor Ontology + Ethics │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY B: EVIDENCE (2 agents) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:b1 LiteratureReviewStrategist Literature search │
│ diverga:b2 EvidenceQualityAppraiser RoB, GRADE appraisal │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY C: DESIGN & META-ANALYSIS (4 agents) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:c1 QuantitativeDesignConsultant Quant design │
│ + Materials + Sampling (absorbed C4, D1) │
│ diverga:c2 QualitativeDesignConsultant Qual design │
│ + Ethnography + Action Research (absorbed H1, H2) │
│ diverga:c3 MixedMethodsDesignConsultant Mixed methods │
│ diverga:c5 MetaAnalysisMaster ⭐ Meta-analysis lead │
│ + Data/Effect/Error/Sensitivity (absorbed C6,C7,B3,E5)│
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY D: DATA COLLECTION (2 agents) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:d2 DataCollectionSpecialist Interview + Observation │
│ (absorbed D3, renamed) │
│ diverga:d4 MeasurementInstrumentDeveloper Instrument dev │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY E: ANALYSIS (3 agents) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:e1 QuantitativeAnalysisGuide Statistical guidance │
│ + Code Gen + Sensitivity (absorbed E4, E5) │
│ diverga:e2 QualitativeCodingSpecialist Qualitative coding │
│ diverga:e3 MixedMethodsIntegration Integration methods │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY F: QUALITY (1 agent) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:f5 HumanizationVerifier Verify humanization │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY G: COMMUNICATION (4 agents) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:g1 JournalMatcher Match journals │
│ diverga:g2 PublicationSpecialist Writing + Review + PreReg│
│ + Quality (absorbed G3, G4, F1, F2, F3) │
│ diverga:g5 AcademicStyleAuditor AI pattern detection │
│ diverga:g6 AcademicStyleHumanizer Humanize AI text │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY I: SYSTEMATIC REVIEW (4 agents) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:i0 ReviewPipelineOrchestrator Pipeline coordination │
│ diverga:i1 PaperRetrievalAgent Multi-database fetch │
│ diverga:i2 ScreeningAssistant AI-PRISMA screening │
│ diverga:i3 RAGBuilder Vector DB + Parallel │
│ (absorbed B5) │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ CATEGORY X: CROSS-CUTTING (1 agent) │
├─────────────────────────────────────────────────────────────────┤
│ diverga:x1 ResearchGuardian Ethics + Bias detection │
│ (absorbed A4, F4) │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ HUMAN CHECKPOINTS │
├─────────────────────────────────────────────────────────────────┤
│ 🔴 REQUIRED (System STOPS): │
│ CP_PARADIGM Research paradigm selection │
│ CP_METHODOLOGY Methodology approval │
│ │
│ 🟠 RECOMMENDED (System PAUSES): │
│ CP_THEORY Theory framework selection │
│ CP_DATA_VALIDATION Data extraction validation │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ MODEL ROUTING │
├─────────────────────────────────────────────────────────────────┤
│ HIGH (Opus): A1,A2,A5,C1,C2,C3,C5,D4,E1,E2,E3,G6,I0 │
│ MEDIUM (Sonnet): B1,B2,D2,G1,G2,G5,X1,I1,I2 │
│ LOW (Haiku): F5,I3 │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│ AUTO-TRIGGER KEYWORDS │
├─────────────────────────────────────────────────────────────────┤
│ "research question", "RQ", "연구 질문" → diverga:a1 │
│ "theoretical framework", "이론적 프레임워크" → diverga:a2 │
│ "critique", "devil's advocate", "반론" → diverga:a2 │
│ "IRB", "ethics", "연구 윤리" → diverga:x1 │
│ "meta-analysis", "메타분석", "효과크기" → diverga:c5 │
│ "systematic review", "PRISMA" → diverga:b1 │
│ "qualitative", "interview", "질적 연구" → diverga:c2 │
│ "ethnography", "action research" → diverga:c2 │
└─────────────────────────────────────────────────────────────────┘
For more info: https://github.com/HosungYou/DivergaUsers can invoke specific agents:
diverga:c5 # Invoke Meta-Analysis Master directly
diverga:a1 # Invoke Research Question Refiner| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | fail→pass | 5,252 | 2,030 | -61% | 1 | 1 | 0% | 867 | 2,229 | +157% | 0 | 0 | — |
case-01 | fail→fail | 18,810 | 18,897 | +0% | 1 | 1 | 0% | 3,535 | 2,029 | -43% | 0 | 0 | — |
case-02 | fail→fail | 12,355 | 24,514 | +98% | 1 | 1 | 0% | 2,032 | 2,215 | +9% | 0 | 0 | — |
case-03 | fail→fail | 10,900 | 16,303 | +50% | 1 | 1 | 0% | 2,037 | 2,085 | +2% | 0 | 0 | — |
case-04 | fail→pass | 9,328 | 2,736 | -71% | 1 | 1 | 0% | 1,549 | 2,396 | +55% | 0 | 0 | — |
case-05 | fail→pass | 4,618 | 1,336 | -71% | 1 | 1 | 0% | 756 | 2,077 | +175% | 0 | 0 | — |
case-06 | fail→pass | 6,845 | 2,788 | -59% | 1 | 1 | 0% | 1,068 | 2,365 | +121% | 0 | 0 | — |
case-07 | fail→pass | 11,925 | 4,270 | -64% | 1 | 1 | 0% | 1,839 | 2,640 | +44% | 0 | 0 | — |
case-08 | fail→pass | 11,095 | 2,470 | -78% | 1 | 1 | 0% | 1,714 | 2,340 | +37% | 0 | 0 | — |
case-09 | fail→pass | 7,358 | 2,314 | -69% | 1 | 1 | 0% | 1,162 | 2,239 | +93% | 0 | 0 | — |
case-10 | fail→pass | 7,788 | 2,336 | -70% | 1 | 1 | 0% | 1,348 | 2,257 | +67% | 0 | 0 | — |
case-11 | fail→pass | 8,241 | 4,591 | -44% | 1 | 1 | 0% | 1,328 | 2,674 | +101% | 0 | 0 | — |
case-12 | fail→pass | 7,231 | 1,753 | -76% | 1 | 1 | 0% | 1,245 | 2,186 | +76% | 0 | 0 | — |
case-13 | fail→pass | 7,216 | 2,986 | -59% | 1 | 1 | 0% | 1,139 | 2,395 | +110% | 0 | 0 | — |
case-14 | fail→pass | 4,980 | 2,459 | -51% | 1 | 1 | 0% | 797 | 2,280 | +186% | 0 | 0 | — |
case-16 | fail→pass | 5,102 | 2,435 | -52% | 1 | 1 | 0% | 843 | 2,345 | +178% | 0 | 0 | — |
case-17 | fail→pass | 11,424 | 2,743 | -76% | 1 | 1 | 0% | 1,818 | 2,416 | +33% | 0 | 0 | — |
case-18 | fail→pass | 5,831 | 1,418 | -76% | 1 | 1 | 0% | 946 | 2,107 | +123% | 0 | 0 | — |
case-19 | fail→pass | 9,172 | 2,362 | -74% | 1 | 1 | 0% | 1,626 | 2,350 | +45% | 0 | 0 | — |
case-20 | pass→pass | 6,800 | 11,614 | +71% | 1 | 1 | 0% | 1,416 | 4,324 | +205% | 0 | 0 | — |
case-21 | pass→pass | 8,221 | 9,790 | +19% | 1 | 1 | 0% | 1,575 | 3,599 | +129% | 0 | 0 | — |
case-22 | pass→pass | 5,695 | 5,673 | -0% | 1 | 1 | 0% | 1,011 | 2,819 | +179% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +73 percentage points is the difference between those two pass rates over the 19 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.