--- name: brycewang-stanford/setup source: https://app.decimal.ai/s/brycewang-stanford-setup@1/SKILL.md source_sha256: 05dbffae9fba --- # /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. ║ ║ ║ ╚══════════════════════════════════════════════════════════════════╝ ``` ```yaml 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: ```yaml 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: ```json { "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`: ```json { "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" } } ``` (merge with existing settings, don't overwrite) **If "Subagents + Classic VS" selected**, add to config: ```json { "agent_teams": { "enabled": false }, "vs_arena": { "enabled": true, "team_size": 3, "cross_critique": false } } ``` **If "Disabled" selected**, add to config: ```json { "agent_teams": { "enabled": false }, "vs_arena": { "enabled": false } } ``` ### Step 4: Generate Configuration & Complete After collecting all preferences, generate `config/diverga-config.json` at `~/.claude/plugins/diverga/config/`: ```json { "version": "12.0.0", "researcher": { "discipline": "Education", "experience": "doctoral_student", "stats_software": ["R", "SPSS"], "db_access": ["Scopus", "ERIC", "Semantic Scholar"] }, "agent_teams": { "enabled": true }, "vs_arena": { "enabled": true, "team_size": 3, "cross_critique": true } } ``` **Experience level mapping:** - "Doctoral Student" → `"doctoral_student"` - "Postdoc / Early Career" → `"early_career"` - "Faculty / Senior Researcher" → `"faculty"` Display completion: ``` ╔══════════════════════════════════════════════════════════════════╗ ║ Diverga v12.0 Setup Complete! ║ ╠══════════════════════════════════════════════════════════════════╣ ║ ║ ║ Profile saved: ║ ║ Discipline: Education ║ ║ Experience: Doctoral Student ║ ║ Stats: R, SPSS ║ ║ Databases: Scopus, ERIC, Semantic Scholar ║ ║ Agent Teams: Enabled ║ ║ VS Arena: Cross-critique enabled ║ ║ ║ ║ 29 agents (24 core + 5 VS Arena) ║ ║ Agent Teams orchestrator — unified parallel and debate workflows║ ║ ║ ║ Quick Start: ║ ║ - Just describe your research in natural language ║ ║ - "I want to conduct a systematic review on AI in education" ║ ║ - Diverga adapts to your profile automatically ║ ║ ║ ║ Commands: ║ ║ - /diverga:help - View all 29 agents ║ ║ - /diverga:memory - Memory system commands ║ ║ - /diverga:setup - Update your profile anytime ║ ║ ║ ╚══════════════════════════════════════════════════════════════════╝ ``` ## Lazy Config (Agent-Level) These settings are NOT asked during setup. Each agent asks when first needed: | Setting | Agent | When Asked | Storage | |---------|-------|------------|---------| | OpenAlex email | G1 (Journal Matcher) | First `journal_search_by_field` call | `.omc/config.json` | | Citation format | G2 (Publication) | First manuscript generation | `researcher.citation_format` in config | | Qualitative software | E2 (Coding) | First qualitative coding task | `researcher.qual_software` in config | **Lazy config prompt template** (for agents to use): When an agent needs a lazy config value that isn't set yet: 1. Explain WHY the information is needed (1 sentence, tied to the current task) 2. Ask via AskUserQuestion with relevant options 3. Save to `config/diverga-config.json` under the `researcher` object 4. Continue with the task — do NOT restart or re-prompt Example (G1 first call): ``` "Journal search uses OpenAlex API. Registering your email enables faster responses (polite pool). This is optional." → AskUserQuestion: [Enter email] [Skip] → If email: save to .omc/config.json ``` ## First-Run Detection When a Diverga plugin session starts: 1. Check if `~/.claude/plugins/diverga/config/diverga-config.json` exists with `researcher` field 2. If missing → display: ``` Welcome to Diverga! Run /diverga:setup to set up your researcher profile (1 minute). ``` 3. Do NOT auto-run — only display the suggestion once. ## Config Schema Reference ```json { "version": "12.0.0", "researcher": { "discipline": "string", "experience": "doctoral_student | early_career | faculty", "stats_software": ["R", "Python", "SPSS", "Stata", "Mplus"], "db_access": ["Scopus", "Web of Science", "PsycINFO", "ERIC", "Semantic Scholar"], "qual_software": "NVivo | ATLAS.ti | MAXQDA | Dedoose | manual", "citation_format": "APA | Chicago | Vancouver | Harvard" }, "agent_teams": { "enabled": "boolean" }, "vs_arena": { "enabled": "boolean", "team_size": "number (default 3)", "cross_critique": "boolean" } } ``` Fields are added incrementally: - `researcher.discipline`, `experience`, `stats_software`, `db_access` — set during `/diverga:setup` - `researcher.qual_software`, `citation_format` — added by lazy config when relevant agent runs - `agent_teams`, `vs_arena` — set during `/diverga:setup` Step 3 ## Error Handling If config directory doesn't exist, create it: ```bash mkdir -p ~/.claude/plugins/diverga/config ```