/setup

name: setup
description: |
  Diverga v12.0 setup wizard. 4-step researcher profile setup.
  Captures discipline, experience, tools, database access, and Agent Teams + VS Arena preference.
  Triggers: setup, configure, 설정, install
version: "12.0.1"

/diverga:setup

**Version**: 12.0.0 **Trigger**: `/diverga:setup`

Description

Diverga setup wizard. 4 steps: Researcher Profile, Tools & Access, Agent Teams + VS Arena, Config Generation. Captures information that genuinely changes agent behavior.

Design Principles

1. **Only ask what agents actually use** — no dead config fields 2. **Lazy config for rare features** — OpenAlex email, citation format, qual software are asked when the relevant agent runs for the first time 3. **Researcher profile is stable** — discipline and tools don't change per project; checkpoints handle project-specific decisions

Workflow

When user invokes `/diverga:setup`, execute this interactive wizard:

Step 0: Project Detection

Check for existing config:

• If `~/.claude/plugins/diverga/config/diverga-config.json` exists with `researcher` field → "Existing profile detected. Update?"
• Otherwise → "New profile setup"

Step 1: Welcome + Researcher Profile

Display welcome message, then ask TWO questions using a single AskUserQuestion call:

╔══════════════════════════════════════════════════════════════════╗
║                    Welcome to Diverga v12.0                     ║
║   AI Research Assistant - 29 Agents (24 core + 5 VS Arena)      ║
╠══════════════════════════════════════════════════════════════════╣
║                                                                  ║
║  Diverga adapts to your background.                             ║
║  A doctoral student gets more scaffolding.                      ║
║  An experienced faculty gets concise recommendations.           ║
║  Your preferred tools determine code and database suggestions.  ║
║                                                                  ║
╚══════════════════════════════════════════════════════════════════╝

questions:
  - question: "What is your primary research discipline?"
    header: "Discipline"
    multiSelect: false
    options:
      - label: "Education"
        description: "Educational Technology, Curriculum, Higher Ed, STEM Ed, etc."
      - label: "Psychology"
        description: "Clinical, Developmental, Social, Cognitive, I/O, etc."
      - label: "Health Sciences"
        description: "Public Health, Nursing, Medicine, Rehabilitation, etc."
      - label: "Social Sciences"
        description: "Sociology, Political Science, Communication, Business, etc."
    # User can also type a custom discipline via "Other"

  - question: "What is your research experience level?"
    header: "Experience"
    multiSelect: false
    options:
      - label: "Doctoral Student"
        description: "Currently pursuing PhD/EdD. More guidance on methodology and analysis."
      - label: "Postdoc / Early Career"
        description: "Completed doctorate. Familiar with research process."
      - label: "Faculty / Senior Researcher"
        description: "Experienced researcher. Concise recommendations preferred."

**How this affects agents:**

• `Doctoral Student` → A1 explains PICO/SPIDER, C1 scaffolds power analysis, E1 adds interpretation guidance
• `Faculty` → Agents skip explanations, go straight to options and trade-offs
• Discipline → G1 prioritizes field-specific journals, I1 leads with field-appropriate databases

Step 2: Tools & Institutional Access

Ask TWO questions using a single AskUserQuestion call:

questions:
  - question: "Which statistical software do you use?"
    header: "Stats Tools"
    multiSelect: true
    options:
      - label: "R"
        description: "tidyverse, lavaan, lme4, etc."
      - label: "Python"
        description: "pandas, statsmodels, scikit-learn, etc."
      - label: "SPSS"
        description: "IBM SPSS Statistics"
      - label: "Stata"
        description: "Stata/SE or Stata/MP"

  - question: "Which academic databases can you access?"
    header: "DB Access"
    multiSelect: true
    options:
      - label: "Scopus"
        description: "Elsevier Scopus (institutional subscription)"
      - label: "Web of Science"
        description: "Clarivate WoS (institutional subscription)"
      - label: "PsycINFO"
        description: "APA PsycINFO (institutional subscription)"
      - label: "ERIC + Semantic Scholar (free)"
        description: "Always available, no subscription needed"

**How this affects agents:**

• Stats tools → E1 generates code ONLY in selected languages (no more 4-language output)
• DB access → I1/I0 recommends only accessible databases at `SCH_DATABASE_SELECTION` checkpoint

Step 3: Agent Teams and VS Arena

question: "Configure multi-agent collaboration mode"
header: "Agent Teams"
options:
  - label: "Agent Teams + VS Arena (Recommended)"
    description: "Full parallel execution with inter-agent communication. VS Arena debates use real cross-critique between personas. Requires Claude Code v2.1.32+. Higher token usage."
  - label: "Subagents + Classic VS (Default)"
    description: "Agents run as subagents. VS Arena generates options from single agent (no cross-critique). Lower cost."
  - label: "Disabled"
    description: "No multi-agent features. Single agent execution only."

**If "Agent Teams + VS Arena" selected**, add to config:

{
  "agent_teams": { "enabled": true },
  "vs_arena": { "enabled": true, "team_size": 3, "cross_critique": true }
}

Also set `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1` in user's `settings.json` env: 1. Write to `~/.claude/settings.json`:

   { "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" } }

(merge with existing settings, don't overwrite)

**If "Subagents + Classic VS" selected**, add to config:

{
  "agent_teams": { "enabled": false },
  "vs_arena": { "enabled": true, "team_size": 3, "cross_