farmwork 1.0.2 → 1.4.0

package/src/init.js

@@ -324,6 +324,8 @@ export async function init(options) {
         fn: async () => {
           await fs.ensureDir(path.join(cwd, "_AUDIT"));
           await fs.ensureDir(path.join(cwd, "_PLANS"));
+          await fs.ensureDir(path.join(cwd, "_RESEARCH"));
+          await fs.ensureDir(path.join(cwd, "_OFFICE"));
           await fs.ensureDir(path.join(cwd, ".claude", "commands"));
           await fs.ensureDir(path.join(cwd, ".claude", "agents"));
         },
@@ -337,6 +339,10 @@ export async function init(options) {
         name: "Creating audit documents",
         fn: () => createAuditDocs(cwd, answers),
       },
+      {
+        name: "Setting up office",
+        fn: () => createOfficeDocs(cwd, answers),
+      },
       { name: "Laying out justfile", fn: () => createJustfile(cwd, answers) },
       { name: "Training agents", fn: () => createAgents(cwd, answers) },
       { name: "Setting up commands", fn: () => createCommands(cwd, answers) },
@@ -432,6 +438,8 @@ export async function init(options) {
       [
         "_AUDIT/",
         "_PLANS/",
+        "_RESEARCH/",
+        "_OFFICE/",
         ".claude/commands/",
         ".claude/agents/",
         "CLAUDE.md",
@@ -556,6 +564,37 @@ Run these in order for a complete development cycle:
 
 ---
 
+### Idea Phrases (Pre-Plan Stage)
+
+| Phrase | Action |
+|--------|--------|
+| **I have an idea for...** | Add new idea to \`_AUDIT/GARDEN.md\` (title, description, bullets) |
+| **let's plan this idea...** | Graduate idea from GARDEN → create plan in \`_PLANS/\` |
+| **I dont want to do this idea...** | Reject idea → move from GARDEN to COMPOST |
+| **remove this feature...** | Archive feature idea to COMPOST |
+| **compost this...** | Move idea from GARDEN to COMPOST |
+| **water the garden** | Generate 10 new ideas based on existing GARDEN and COMPOST |
+
+---
+
+### Research Phrases (Pre-Plan Stage)
+
+| Phrase | Action |
+|--------|--------|
+| **let's research...** | Create or update research document in \`_RESEARCH/\` |
+| **update research on...** | Update existing research document with fresh findings |
+| **show research on...** | Display summary of existing research document |
+
+---
+
+### Office Phrases (Product Strategy & UX)
+
+| Phrase | Action |
+|--------|--------|
+| **go to production** | UX production check: update ONBOARDING.md, USER_GUIDE.md, audit CORE_LOOP.md changes |
+
+---
+
 ### Farmwork Phrase Details
 
 **open the farm**
@@ -585,6 +624,125 @@ Runs all inspection agents in parallel, then dry run quality gates. No push.
 
 ---
 
+### Idea Phrase Details
+
+**I have an idea for...**
+1. Launch \`idea-gardener\` agent
+2. Parse idea title from user input
+3. Ask for short description and bullet points
+4. Add to \`_AUDIT/GARDEN.md\` under ## Ideas section
+
+**let's plan this idea...**
+1. Launch \`idea-gardener\` agent
+2. Find the idea in GARDEN.md
+3. Create plan in \`_PLANS/\` using plan mode
+4. Move to "Graduated to Plans" table
+5. Remove from ## Ideas section
+
+**compost this...** / **I dont want to do this idea...**
+1. Launch \`idea-gardener\` agent
+2. Find idea in GARDEN.md (or accept new rejection)
+3. Ask for rejection reason
+4. Add to \`_AUDIT/COMPOST.md\` with reason
+5. Remove from GARDEN.md if it was there
+
+**water the garden**
+1. Launch \`idea-gardener\` agent
+2. Read \`_AUDIT/GARDEN.md\` to understand existing ideas and themes
+3. Read \`_AUDIT/COMPOST.md\` to understand what didn't work and why
+4. Generate 10 new, creative ideas that:
+   - Build on or extend existing garden ideas
+   - Avoid patterns that led to composted ideas
+   - Consider the project's direction and goals
+5. Present ideas as a numbered list with title and one-line description
+6. Ask user: "Which ideas would you like to plant? (enter numbers, e.g., 1, 3, 5)"
+7. For selected ideas, add each to GARDEN.md with today's planted date
+
+---
+
+### Research Phrase Details
+
+**let's research...**
+1. Launch \`researcher\` agent
+2. Parse research topic from user input
+3. Check for existing research in \`_RESEARCH/\`
+4. Spawn parallel subagents for:
+   - Documentation finder (official docs, API refs)
+   - Security researcher (CVEs, known issues)
+   - Tech stack analyzer (dependencies, compatibility)
+   - Community researcher (gotchas, discussions)
+5. If ref.tools MCP available: Use \`mcp__Ref__ref_search_documentation\` for docs lookup
+6. Consolidate findings into \`_RESEARCH/[TOPIC_NAME].md\`
+7. Display summary and suggest next steps
+
+**update research on...**
+1. Launch \`researcher\` agent
+2. Find existing research document in \`_RESEARCH/\`
+3. Run targeted research refresh on specified areas
+4. Merge new findings, mark outdated info with strikethrough
+5. Update research history and Last Researched date
+
+**show research on...**
+1. Find research document in \`_RESEARCH/\`
+2. Display summary of key findings, risks, confidence level
+3. Suggest refresh if research is aging (15+ days) or stale (30+ days)
+
+---
+
+### Office Phrase Details
+
+**go to production**
+
+Production readiness check from a user experience perspective. Separate from "close the farm" (which handles code quality/push).
+
+1. **Update ONBOARDING.md**
+   - Spawn \`onboarding-agent\` to scan for onboarding elements
+   - Check for incomplete or missing onboarding flows
+   - Update Last Updated date
+   - Add changelog entry if changes found
+
+2. **Update USER_GUIDE.md**
+   - Spawn \`user-guide-agent\` to scan for undocumented features
+   - Check for placeholder text or incomplete sections
+   - Update feature count
+   - Add changelog entry if changes found
+
+3. **Audit CORE_LOOP.md**
+   - Spawn \`strategy-agent\` to check if CORE_LOOP.md has changed since last production check
+   - If changed, add audit trail entry to Strategy Changelog
+   - Report strategy evolution summary
+
+4. **Generate Production Readiness Report**
+   \`\`\`
+   ## Production Readiness: User Experience
+
+   ### Strategy Status
+   - Last Updated: YYYY-MM-DD
+   - Changes Since Last Deploy: Yes/No
+   - Confidence: High/Medium/Low
+
+   ### Onboarding Status
+   - Elements: X documented
+   - Gaps: X identified
+   - Empty States: X complete
+
+   ### Documentation Status
+   - Features Documented: X
+   - Quick Start: Complete/Incomplete
+   - FAQ: X entries
+
+   ### Recommendation
+   [Ready for production / Needs attention: ...]
+   \`\`\`
+
+5. **Ask for Confirmation**
+   - "UX production check complete. Ready to proceed with deployment?"
+   - Wait for user confirmation before any further action
+
+**Note:** This phrase focuses on UX readiness. Use "close the farm" for code quality gates and pushing to remote.
+
+---
+
 ## Plan Mode Protocol
 
 **CRITICAL**: When Claude enters Plan Mode, ALL plans MUST:
@@ -650,9 +808,11 @@ async function createFarmhouseMd(cwd, answers) {
 
 | Metric | Count |
 |--------|-------|
-| Commands | 1 |
-| Agents | 10 |
-| Justfile Recipes | 10 |
+| Commands | 2 |
+| Agents | 15 |
+| Office Docs | 3 |
+| Research Docs | 0 |
+| Justfile Recipes | 11 |
 | Unit Tests | 0 |
 | E2E Tests | 0 |
 | Completed Issues | 0 |
@@ -670,6 +830,7 @@ All Claude Code commands and agents are documented, phrase triggers are tested a
 | Command | Description |
 |---------|-------------|
 | \`/push\` | Clean, lint, test, build, commit, push, update metrics |
+| \`/office\` | Interactive strategy and UX command - updates CORE_LOOP, ONBOARDING, USER_GUIDE |
 
 ---
 
@@ -687,6 +848,11 @@ All Claude Code commands and agents are documented, phrase triggers are tested a
 | \`code-cleaner\` | Remove comments and console.logs |
 | \`i18n-locale-translator\` | Translate UI text to locales |
 | \`storybook-maintainer\` | Create/update Storybook stories |
+| \`idea-gardener\` | Manage Idea Garden and Compost |
+| \`researcher\` | Systematic research before planning |
+| \`strategy-agent\` | Analyze core loop strategy (what/stopping/why) |
+| \`onboarding-agent\` | Document onboarding elements (tours, tooltips, modals) |
+| \`user-guide-agent\` | Create feature documentation for help docs |
 
 ---
 
@@ -708,6 +874,28 @@ All Claude Code commands and agents are documented, phrase triggers are tested a
 | \`make a plan for...\` | Create plan in _PLANS/ |
 | \`let's implement...\` | Load plan, create Epic |
 
+### Idea Phrases
+
+| Phrase | Action |
+|--------|--------|
+| \`I have an idea for...\` | Add idea to GARDEN.md |
+| \`let's plan this idea...\` | Graduate idea to _PLANS/ |
+| \`compost this...\` | Move idea to COMPOST.md |
+
+### Research Phrases
+
+| Phrase | Action |
+|--------|--------|
+| \`let's research...\` | Research topic, save to _RESEARCH/ |
+| \`update research on...\` | Refresh existing research |
+| \`show research on...\` | Display research summary |
+
+### Office Phrases
+
+| Phrase | Action |
+|--------|--------|
+| \`go to production\` | UX production check: update _OFFICE/ docs |
+
 ---
 
 ## Issue Tracking (\`.beads/\`)
@@ -810,6 +998,312 @@ _None currently_
 
     await fs.writeFile(path.join(cwd, "_AUDIT", audit.name), content);
   }
+
+  // Create GARDEN.md (Idea nursery - custom format)
+  const gardenContent = `# Idea Garden
+
+> Nursery for new ideas and concepts. The pre-plan creative thinking stage.
+> Ideas older than 60 days without action will naturally compost during "open the farm".
+
+**Last Updated:** ${today}
+**Active Ideas:** 0
+**Wilting Ideas:** 0
+
+---
+
+## How to Use
+
+| Phrase | Action |
+|--------|--------|
+| \`I have an idea for...\` | Plant a new idea here |
+| \`let's plan this idea...\` | Graduate idea to _PLANS/ |
+| \`compost this...\` | Reject idea, move to COMPOST |
+
+---
+
+## Idea Lifecycle
+
+Ideas have a natural lifecycle:
+- **Fresh** (0-44 days) - New ideas, ready to be developed
+- **Wilting** (45-60 days) - Ideas aging without action, marked with ⚠️
+- **Composted** (60+ days) - Auto-moved to COMPOST during "open the farm"
+
+---
+
+## Ideas
+
+_No ideas planted yet. Start with "I have an idea for..."_
+
+<!-- Idea format:
+### [Idea Title]
+**Planted:** YYYY-MM-DD
+[Short description]
+- Bullet point 1
+- Bullet point 2
+-->
+
+---
+
+## Graduated to Plans
+
+| Idea | Plan | Date |
+|------|------|------|
+
+---
+
+## Implemented
+
+| Idea | Plan | Completed |
+|------|------|-----------|
+`;
+
+  await fs.writeFile(path.join(cwd, "_AUDIT", "GARDEN.md"), gardenContent);
+
+  // Create COMPOST.md (Rejected ideas archive - custom format)
+  const compostContent = `# Idea Compost
+
+> Archive of rejected ideas. Reference to avoid re-proposing and remember why we didn't pursue something.
+> Ideas that age 60+ days in the Garden are automatically composted during "open the farm".
+
+**Last Updated:** ${today}
+**Composted Ideas:** 0
+**Auto-Composted:** 0
+
+---
+
+## How to Use
+
+| Phrase | Action |
+|--------|--------|
+| \`I dont want to do this idea...\` | Reject an idea |
+| \`remove this feature...\` | Archive a feature idea |
+| \`compost this...\` | Move idea from GARDEN here |
+
+---
+
+## Composted Ideas
+
+_No composted ideas yet._
+
+<!-- Composted idea format:
+### [Idea Title]
+**Composted:** YYYY-MM-DD
+**Reason:** [User's reason OR "Auto-composted: aged 60+ days without action"]
+[Original description if available]
+-->
+`;
+
+  await fs.writeFile(path.join(cwd, "_AUDIT", "COMPOST.md"), compostContent);
+}
+
+async function createOfficeDocs(cwd, answers) {
+  const today = new Date().toISOString().split("T")[0];
+
+  // Create CORE_LOOP.md - Strategy document
+  const coreLoopContent = `# Core Loop Strategy
+
+> Treat your product like a game. Define what users are doing, what's stopping them, and why they're doing it.
+> This is a living strategy document - update it as your understanding evolves.
+
+**Last Updated:** ${today}
+**Status:** Initial setup
+**Confidence:** Low
+
+---
+
+## The Three Questions
+
+### 1. What are they doing?
+_What is the primary action or loop your users engage in?_
+
+**Current Understanding:**
+[Describe the core user action/loop]
+
+### 2. What's stopping them?
+_What friction, obstacles, or pain points prevent users from succeeding?_
+
+**Current Blockers:**
+- [Blocker 1]
+- [Blocker 2]
+
+### 3. Why are they doing it?
+_What motivates users? What's the deeper goal or reward?_
+
+**User Motivation:**
+[Describe the underlying motivation]
+
+---
+
+## Core Loop Diagram
+
+\`\`\`
+[Entry Point] → [Core Action] → [Reward/Feedback] → [Loop Back]
+\`\`\`
+
+---
+
+## Strategy Changelog
+
+| Date | Change | Previous | Reason |
+|------|--------|----------|--------|
+| ${today} | Initial strategy setup | - | Created via farmwork init |
+
+---
+
+## Related Documents
+
+- [ONBOARDING.md](./_OFFICE/ONBOARDING.md) - First-time user experience
+- [USER_GUIDE.md](./_OFFICE/USER_GUIDE.md) - Feature documentation
+`;
+
+  await fs.writeFile(path.join(cwd, "_OFFICE", "CORE_LOOP.md"), coreLoopContent);
+
+  // Create ONBOARDING.md - Onboarding tracker
+  const onboardingContent = `# User Onboarding
+
+> Living document for first-time user experience: tours, popups, modals, tooltips, and progressive disclosure.
+> Track what users see when they first use your product.
+
+**Last Updated:** ${today}
+**Status:** Initial setup
+**Onboarding Steps:** 0
+
+---
+
+## Onboarding Flow Overview
+
+\`\`\`
+[Landing] → [Signup] → [Welcome] → [First Action] → [Success Moment]
+\`\`\`
+
+---
+
+## Onboarding Elements
+
+### Welcome Experience
+_What does the user see immediately after signup/first visit?_
+
+| Element | Type | Content | Status |
+|---------|------|---------|--------|
+| | | | |
+
+### Guided Tours
+_Step-by-step tours that walk users through features_
+
+| Tour Name | Steps | Trigger | Status |
+|-----------|-------|---------|--------|
+| | | | |
+
+### Tooltips & Hints
+_Contextual help that appears on specific elements_
+
+| Element | Tooltip Text | Trigger | Status |
+|---------|--------------|---------|--------|
+| | | | |
+
+### Modals & Popups
+_Modal dialogs that appear during onboarding_
+
+| Modal Name | Purpose | Trigger | Status |
+|------------|---------|---------|--------|
+| | | | |
+
+### Empty States
+_What users see before they have data_
+
+| Screen | Empty State Message | CTA | Status |
+|--------|---------------------|-----|--------|
+| | | | |
+
+---
+
+## Success Metrics
+
+| Metric | Current | Target |
+|--------|---------|--------|
+| Onboarding completion rate | TBD | 80% |
+| Time to first value | TBD | < 5 min |
+| Drop-off points | TBD | Identify |
+
+---
+
+## Changelog
+
+| Date | Change | Reason |
+|------|--------|--------|
+| ${today} | Initial onboarding setup | Created via farmwork init |
+`;
+
+  await fs.writeFile(path.join(cwd, "_OFFICE", "ONBOARDING.md"), onboardingContent);
+
+  // Create USER_GUIDE.md - Feature documentation
+  const userGuideContent = `# User Guide
+
+> Living documentation for features and how to use them.
+> Grows over time to eventually become help docs.
+> Each feature gets a short block with bullet list instructions.
+
+**Last Updated:** ${today}
+**Status:** Initial setup
+**Features Documented:** 0
+
+---
+
+## Quick Start
+
+_Minimal steps to get started with the product_
+
+1. [First step]
+2. [Second step]
+3. [Third step]
+
+---
+
+## Features
+
+<!-- Feature Template:
+### Feature Name
+Brief description of what this feature does.
+
+**How to use:**
+- Step 1
+- Step 2
+- Step 3
+
+**Tips:**
+- Helpful tip
+
+**Related:** [Link to related feature]
+-->
+
+_No features documented yet. Run \`/office\` to add features._
+
+---
+
+## Keyboard Shortcuts
+
+| Shortcut | Action |
+|----------|--------|
+| | |
+
+---
+
+## FAQ
+
+### Common Questions
+
+_No FAQs yet._
+
+---
+
+## Changelog
+
+| Date | Change |
+|------|--------|
+| ${today} | Initial user guide setup |
+`;
+
+  await fs.writeFile(path.join(cwd, "_OFFICE", "USER_GUIDE.md"), userGuideContent);
 }
 
 async function createJustfile(cwd, answers) {
@@ -851,6 +1345,10 @@ audit:
 plans:
     @echo "{{project_root}}/_PLANS" && cd {{project_root}}/_PLANS
 
+# Go to research folder
+research:
+    @echo "{{project_root}}/_RESEARCH" && cd {{project_root}}/_RESEARCH
+
 # Go to commands folder
 commands:
     @echo "{{project_root}}/.claude/commands" && cd {{project_root}}/.claude/commands
@@ -901,7 +1399,7 @@ async function createAgents(cwd, answers) {
 name: the-farmer
 description: Audit and update FARMHOUSE.md with current project metrics
 tools: Read, Grep, Glob, Edit, Bash
-model: haiku
+model: opus
 ---
 
 # The Farmer Agent
@@ -910,12 +1408,52 @@ Maintains \`_AUDIT/FARMHOUSE.md\` - the living document tracking all systems and
 
 ## Instructions
 
+### Step 1: Gather Metrics
 1. Count commands: \`ls -1 .claude/commands/*.md | wc -l\`
 2. Count agents: \`ls -1 .claude/agents/*.md | wc -l\`
 3. Count tests: \`find . -name "*.test.*" | wc -l\`
 4. Count completed issues: \`bd list --status closed | wc -l\`
-5. Update FARMHOUSE.md with fresh metrics
-6. Update score based on completeness
+
+### Step 2: Tend the Idea Garden
+Read \`_AUDIT/GARDEN.md\` and check the age of each idea:
+
+1. Parse each idea's \`**Planted:**\` date
+2. Calculate age: today - planted date (in days)
+3. For ideas **45-60 days old** (Wilting):
+   - Add \`⚠️ WILTING\` after the idea title
+   - Report these ideas in the audit summary
+4. For ideas **over 60 days old** (Composted):
+   - Move to \`_AUDIT/COMPOST.md\` with format:
+     \`\`\`markdown
+     ### [Idea Title]
+     **Composted:** YYYY-MM-DD
+     **Reason:** Auto-composted: aged 60+ days without action
+     [Original description]
+     \`\`\`
+   - Remove from GARDEN.md
+   - Update counts in both files
+5. Update GARDEN.md header:
+   - **Active Ideas:** (count of non-wilting ideas)
+   - **Wilting Ideas:** (count of 45-60 day old ideas)
+   - **Last Updated:** today's date
+
+### Step 3: Check Research Freshness
+Read all documents in \`_RESEARCH/\` and check their age:
+
+1. Parse each document's \`**Last Researched:**\` date
+2. Calculate age: today - last researched date (in days)
+3. For research **15-30 days old** (Aging):
+   - Update status to "Aging" in the document
+   - Report these in the audit summary
+4. For research **over 30 days old** (Stale):
+   - Update status to "Stale" in the document
+   - Report these as needing refresh before use in plans
+5. Count total research documents for FARMHOUSE metrics
+
+### Step 4: Update FARMHOUSE.md
+1. Update metrics table (including Research Docs count)
+2. Update score based on completeness
+3. Add audit history entry
 
 ## Output Format
 
@@ -925,9 +1463,20 @@ Maintains \`_AUDIT/FARMHOUSE.md\` - the living document tracking all systems and
 ### Metrics Updated
 - Commands: X total
 - Agents: X total
+- Research Docs: X total
 - Tests: X files
 - Completed Issues: X total
 
+### Idea Garden
+- Active Ideas: X
+- Wilting Ideas: X (list titles if any)
+- Auto-Composted: X (list titles if any)
+
+### Research Library
+- Fresh: X documents
+- Aging: X documents (list titles if any)
+- Stale: X documents (list titles if any)
+
 ### Score: X/10
 \`\`\`
 `,
@@ -935,7 +1484,7 @@ Maintains \`_AUDIT/FARMHOUSE.md\` - the living document tracking all systems and
 name: code-reviewer
 description: Review code for quality, security, and maintainability
 tools: Read, Grep, Glob, Bash
-model: sonnet
+model: opus
 ---
 
 # Code Reviewer Agent
@@ -952,7 +1501,7 @@ Reports findings with severity (CRITICAL, HIGH, MEDIUM, LOW) and remediation ste
 name: security-auditor
 description: OWASP security vulnerability scanning
 tools: Read, Grep, Glob, Edit
-model: haiku
+model: opus
 ---
 
 # Security Auditor Agent
@@ -971,7 +1520,7 @@ Updates \`_AUDIT/SECURITY.md\` with results.
 name: performance-auditor
 description: Find memory leaks, unnecessary re-renders, and anti-patterns
 tools: Read, Grep, Glob, Edit
-model: haiku
+model: opus
 ---
 
 # Performance Auditor Agent
@@ -990,7 +1539,7 @@ Updates \`_AUDIT/PERFORMANCE.md\` with results.
 name: code-smell-auditor
 description: Detect DRY violations, complexity issues, naming problems, and technical debt
 tools: Read, Grep, Glob, Edit
-model: haiku
+model: opus
 ---
 
 # Code Smell Auditor Agent
@@ -1009,7 +1558,7 @@ Updates \`_AUDIT/CODE_QUALITY.md\` with results.
 name: accessibility-auditor
 description: WCAG 2.1 accessibility auditing for React/Next.js applications
 tools: Read, Grep, Glob, Edit
-model: haiku
+model: opus
 ---
 
 # Accessibility Auditor Agent
@@ -1029,7 +1578,7 @@ Updates \`_AUDIT/ACCESSIBILITY.md\` with results.
 name: unused-code-cleaner
 description: Detect and remove unused code (imports, functions, variables)
 tools: Read, Write, Edit, Bash, Grep, Glob
-model: haiku
+model: opus
 ---
 
 # Unused Code Cleaner Agent
@@ -1048,7 +1597,7 @@ Use after refactoring, when removing features, or before production deployment.
 name: code-cleaner
 description: Fast removal of comments, console.logs, and debug code while preserving JSDoc
 tools: Read, Edit, Glob, Grep
-model: haiku
+model: opus
 ---
 
 # Code Cleaner Agent
@@ -1068,7 +1617,7 @@ Fast cleanup of TypeScript/JavaScript files:
 name: i18n-locale-translator
 description: Translate UI text content into English (en) and Japanese (jp) using i18n locale system
 tools: Read, Write, Edit, Glob, Grep
-model: sonnet
+model: opus
 ---
 
 # i18n Locale Translator Agent
@@ -1085,7 +1634,7 @@ Use when adding new features or internationalizing existing hardcoded text.
 name: storybook-maintainer
 description: Create and update Storybook stories for UI components
 tools: Read, Write, Edit, Glob, Grep
-model: haiku
+model: opus
 ---
 
 # Storybook Maintainer Agent
@@ -1097,6 +1646,530 @@ Manages Storybook stories for UI components:
 - Add controls for interactive props
 
 Use when adding new components or when existing components change significantly.
+`,
+    "idea-gardener.md": `---
+name: idea-gardener
+description: Manage the Idea Garden and Compost - add, graduate, reject, or generate ideas
+tools: Read, Edit, Glob, Grep
+model: opus
+---
+
+# Idea Gardener Agent
+
+Manages \`_AUDIT/GARDEN.md\` and \`_AUDIT/COMPOST.md\` for idea lifecycle tracking.
+
+## Commands
+
+### Plant an Idea (from "I have an idea for...")
+1. Parse the idea title from user input
+2. Ask user for short description and key bullet points
+3. Add to GARDEN.md under ## Ideas section with format:
+   \`\`\`markdown
+   ### [Idea Title]
+   **Planted:** YYYY-MM-DD
+   [Short description]
+   - Bullet point 1
+   - Bullet point 2
+   \`\`\`
+4. Update the "Active Ideas" count in the header
+5. Update "Last Updated" date
+
+**IMPORTANT:** Always include the **Planted:** date using today's date (YYYY-MM-DD format).
+
+### Graduate an Idea (from "let's plan this idea...")
+1. Find idea in GARDEN.md
+2. Create plan file in _PLANS/ using plan mode
+3. Move idea to "Graduated to Plans" table with date and plan link
+4. Remove from ## Ideas section
+5. Update "Active Ideas" count
+
+### Compost an Idea (from "compost this..." / "I dont want...")
+1. Find idea in GARDEN.md (or accept new rejection)
+2. Ask for rejection reason
+3. Add to COMPOST.md with format:
+   \`\`\`markdown
+   ### [Idea Title]
+   **Composted:** YYYY-MM-DD
+   **Reason:** [User's reason]
+   [Original description if available]
+   \`\`\`
+4. Remove from GARDEN.md if it was there
+5. Update counts in both files
+
+### Water the Garden (from "water the garden")
+Generate fresh ideas based on the project context:
+
+1. **Read Context:**
+   - Read \`_AUDIT/GARDEN.md\` - understand existing ideas, themes, what's being explored
+   - Read \`_AUDIT/COMPOST.md\` - understand what was rejected and why (avoid these patterns)
+   - Read \`CLAUDE.md\` - understand the project's purpose and configuration
+
+2. **Generate 10 Ideas:**
+   Think creatively about ideas that:
+   - Extend or complement existing garden ideas
+   - Fill gaps in current thinking
+   - Avoid patterns that led to rejected/composted ideas
+   - Align with the project's goals and tech stack
+   - Range from small enhancements to ambitious features
+
+3. **Present Ideas:**
+   Display as a numbered list:
+   \`\`\`
+   ## Fresh Ideas for Your Garden
+
+   1. **[Idea Title]** - One-line description
+   2. **[Idea Title]** - One-line description
+   ... (10 total)
+
+   Which ideas would you like to plant? (enter numbers, e.g., 1, 3, 5)
+   \`\`\`
+
+4. **Plant Selected Ideas:**
+   For each selected number, add to GARDEN.md with:
+   - Title from the list
+   - Today's date as **Planted:** date
+   - The one-line description expanded slightly
+   - 2-3 bullet points about potential implementation
+
+## Output Format
+Confirm action taken and show updated file section.
+`,
+    "researcher.md": `---
+name: researcher
+description: Systematic research agent - gathers documentation, risks, security concerns, and implementation insights
+tools: Read, Edit, Glob, Grep, Bash, WebFetch, Task
+model: opus
+---
+
+# Researcher Agent
+
+Conducts systematic research on features, technologies, and concepts before planning.
+Creates and maintains living research documents in \`_RESEARCH/\`.
+
+## Core Capabilities
+
+1. **Parallel Research Spawning** - Spawns focused subagents for different research areas
+2. **Documentation Discovery** - Finds official docs, API references, tutorials
+3. **Security Analysis** - Identifies CVEs, known vulnerabilities, security best practices
+4. **Tech Stack Analysis** - Analyzes dependencies, compatibility, bundle size
+5. **Community Insights** - Gathers gotchas, common issues, best practices from community
+6. **MCP Integration** - Uses ref.tools MCP when available for enhanced documentation access
+
+## Instructions
+
+### Step 1: Parse Research Request
+1. Extract the research topic from user input after "let's research..."
+2. Normalize topic name to SCREAMING_SNAKE_CASE for filename
+3. Check if \`_RESEARCH/[TOPIC_NAME].md\` already exists
+
+### Step 2: Spawn Parallel Research Agents
+Create focused subtasks for parallel execution using the Task tool:
+
+**Documentation Research Task:**
+- Find official documentation sites
+- Identify API references and getting started guides
+- Locate migration guides if applicable
+- If ref.tools MCP available: Query \`mcp__Ref__ref_search_documentation\` for relevant docs
+
+**Security Research Task:**
+- Search for known CVEs related to the topic
+- Find security advisories
+- Identify authentication/authorization concerns
+- Research data handling best practices
+- Check for dependency vulnerabilities
+
+**Tech Stack Research Task:**
+- Identify required dependencies
+- Check Node.js/browser compatibility
+- Analyze bundle size implications
+- Find TypeScript type definitions
+- Check for ESM/CJS compatibility
+
+**Community Research Task:**
+- Search GitHub issues for common problems
+- Find Stack Overflow discussions
+- Identify known gotchas and edge cases
+- Gather migration experiences
+- Find performance optimization tips
+
+### Step 3: Consolidate Findings
+1. Wait for all parallel research tasks to complete
+2. Merge findings into structured research document
+3. Identify conflicts or contradictions between sources
+4. Assign confidence levels based on source quality
+5. Highlight critical risks that require attention
+
+### Step 4: Create/Update Research Document
+
+**If new research:**
+Create \`_RESEARCH/[TOPIC_NAME].md\` with this format:
+
+\`\`\`markdown
+# Research: [Topic Name]
+
+> Systematic research findings for informed decision-making.
+> This is a living document - updated periodically as new information emerges.
+
+**Created:** YYYY-MM-DD
+**Last Researched:** YYYY-MM-DD
+**Status:** Fresh
+**Confidence:** High | Medium | Low
+
+---
+
+## Summary
+
+[2-3 sentence executive summary of key findings]
+
+---
+
+## Official Documentation
+
+| Resource | URL | Notes |
+|----------|-----|-------|
+| [Doc Name] | [URL] | [Key insight] |
+
+---
+
+## Tech Stack Analysis
+
+### Dependencies
+- **Package Name** - version X.X.X - [purpose/notes]
+
+### Compatibility
+| Environment | Status | Notes |
+|-------------|--------|-------|
+| Node.js | vX.X+ | [notes] |
+| Browser | [support] | [notes] |
+
+### Bundle Size / Performance
+[Analysis of size and performance implications]
+
+---
+
+## Security Concerns
+
+### Known Vulnerabilities
+| CVE/Issue | Severity | Status | Mitigation |
+|-----------|----------|--------|------------|
+| [CVE-ID] | High/Med/Low | Fixed/Open | [action] |
+
+### Security Best Practices
+- [Practice 1]
+- [Practice 2]
+
+---
+
+## Risks & Gotchas
+
+### Common Pitfalls
+1. **[Pitfall Name]** - [Description and how to avoid]
+
+### Breaking Changes
+| Version | Change | Impact |
+|---------|--------|--------|
+| [ver] | [change] | [impact] |
+
+### Edge Cases
+- [Edge case 1]
+- [Edge case 2]
+
+---
+
+## Community Insights
+
+### GitHub Issues / Discussions
+| Issue | Topic | Resolution |
+|-------|-------|------------|
+| [#123] | [topic] | [resolution] |
+
+### Stack Overflow / Forums
+- [Key insight from community]
+
+---
+
+## Implementation Recommendations
+
+### Recommended Approach
+[Based on research, the recommended approach is...]
+
+### Alternatives Considered
+| Approach | Pros | Cons |
+|----------|------|------|
+| [Alt 1] | [pros] | [cons] |
+
+---
+
+## Related Research
+
+- [Link to related _RESEARCH/ document]
+- [Link to relevant _PLANS/ document]
+
+---
+
+## Research History
+
+| Date | Researcher | Areas Updated |
+|------|------------|---------------|
+| YYYY-MM-DD | researcher agent | Initial research |
+\`\`\`
+
+**If updating existing research:**
+1. Read existing document
+2. Merge new findings with existing content
+3. Mark outdated information with ~~strikethrough~~
+4. Update Last Researched date
+5. Update Status based on age (Fresh: 0-14d, Aging: 15-30d, Stale: 30+d)
+6. Add entry to Research History table
+
+### Step 5: Integration Check
+1. Check for related ideas in \`_AUDIT/GARDEN.md\`
+2. Check for existing plans in \`_PLANS/\`
+3. Add cross-references to Related Research section
+4. Suggest next steps (plan creation, more research, etc.)
+
+## Staleness Detection
+
+Research document status:
+- **Fresh** (0-14 days) - Research is current and reliable
+- **Aging** (15-30 days) - Consider refreshing for major decisions
+- **Stale** (30+ days) - Recommend updating before using for plans
+
+## Output Format
+
+After research completion, display:
+
+\`\`\`
+## Research Complete: [Topic Name]
+
+### Key Findings
+- [Most important finding 1]
+- [Most important finding 2]
+- [Most important finding 3]
+
+### Critical Risks
+- [Risk 1 if any]
+- [Risk 2 if any]
+
+### Confidence: [High/Medium/Low]
+
+Research document saved to: _RESEARCH/[TOPIC_NAME].md
+
+Next steps:
+- [ ] Review full research document
+- [ ] "make a plan for..." to create implementation plan
+- [ ] "update research on..." to gather more information
+\`\`\`
+`,
+    "strategy-agent.md": `---
+name: strategy-agent
+description: Analyze and update the core loop strategy - what users do, what stops them, why they do it
+tools: Read, Edit, Glob, Grep, Task
+model: opus
+---
+
+# Strategy Agent
+
+Maintains \`_OFFICE/CORE_LOOP.md\` - the living strategy document for product thinking.
+
+## Core Responsibility
+
+Treat the product like a game and answer three fundamental questions:
+1. **What are they doing?** - The primary user action/loop
+2. **What's stopping them?** - Friction and obstacles
+3. **Why are they doing it?** - Underlying motivation and rewards
+
+## Instructions
+
+### When Invoked via /office
+
+1. Read \`_OFFICE/CORE_LOOP.md\` to understand current strategy
+2. Read the codebase to understand the product (components, routes, features)
+3. Ask the user probing questions:
+   - "What's the main thing users do in your app?"
+   - "Where do users get stuck or confused?"
+   - "What's the 'aha moment' for users?"
+   - "What brings users back?"
+4. Update CORE_LOOP.md with insights
+5. Add entry to Strategy Changelog table
+
+### When Checking for Production ("go to production")
+
+1. Read \`_OFFICE/CORE_LOOP.md\`
+2. Compare current date to Last Updated date
+3. If changed since last production push:
+   - Summarize strategy changes
+   - Add audit entry to changelog
+4. Report strategy status
+
+## Output Format
+
+\`\`\`
+## Strategy Analysis
+
+### Core Loop
+[Describe the identified core loop]
+
+### Key Friction Points
+- [Friction 1]
+- [Friction 2]
+
+### User Motivation
+[Describe why users engage]
+
+### Recommendations
+- [Recommendation 1]
+- [Recommendation 2]
+
+Updated _OFFICE/CORE_LOOP.md
+\`\`\`
+`,
+    "onboarding-agent.md": `---
+name: onboarding-agent
+description: Identify and document onboarding elements - tours, popups, modals, tooltips, empty states
+tools: Read, Edit, Glob, Grep, Task
+model: opus
+---
+
+# Onboarding Agent
+
+Maintains \`_OFFICE/ONBOARDING.md\` - tracking first-time user experience elements.
+
+## Core Responsibility
+
+Identify, document, and track all onboarding-related UI elements:
+- Welcome experiences
+- Guided tours
+- Tooltips and hints
+- Modals and popups
+- Empty states
+- Progressive disclosure
+
+## Instructions
+
+### When Invoked via /office
+
+1. Read \`_OFFICE/ONBOARDING.md\` to understand current state
+2. Scan the codebase for onboarding elements:
+   - Search for: \`tour\`, \`tooltip\`, \`modal\`, \`popup\`, \`hint\`, \`onboarding\`, \`welcome\`, \`empty\`, \`first-time\`
+   - Check for libraries: \`react-joyride\`, \`intro.js\`, \`shepherd.js\`, etc.
+3. Ask the user questions:
+   - "What should users see on their first visit?"
+   - "What's the critical 'aha moment' you want to guide them to?"
+   - "Are there any tours or tooltips currently implemented?"
+4. Document findings in ONBOARDING.md tables
+5. Identify gaps in onboarding coverage
+6. Add entry to Changelog
+
+### When Checking for Production ("go to production")
+
+1. Read \`_OFFICE/ONBOARDING.md\`
+2. Check for incomplete or missing onboarding:
+   - Empty states without content
+   - Key flows without guidance
+   - Tooltips without text
+3. Report onboarding readiness status
+4. Update Last Updated date
+
+## Output Format
+
+\`\`\`
+## Onboarding Analysis
+
+### Found Elements
+- [X] Welcome modal
+- [ ] Guided tour
+- [X] Empty states (3 found)
+
+### Gaps Identified
+- No tour for main feature
+- Missing tooltip on key button
+
+### Recommendations
+- Add 3-step tour for new users
+- Create empty state for dashboard
+
+Updated _OFFICE/ONBOARDING.md
+\`\`\`
+`,
+    "user-guide-agent.md": `---
+name: user-guide-agent
+description: Document features and create user help documentation in bullet list format
+tools: Read, Edit, Glob, Grep, Task
+model: opus
+---
+
+# User Guide Agent
+
+Maintains \`_OFFICE/USER_GUIDE.md\` - living feature documentation that grows into help docs.
+
+## Core Responsibility
+
+Create and maintain user-facing documentation for features:
+- Short, scannable feature descriptions
+- Step-by-step bullet list instructions
+- Helpful tips and related features
+- Keyboard shortcuts
+- FAQ entries
+
+## Instructions
+
+### When Invoked via /office
+
+1. Read \`_OFFICE/USER_GUIDE.md\` to understand current documentation
+2. Scan the codebase for features:
+   - Identify routes and pages
+   - Find user-facing components
+   - Look for keyboard event handlers
+   - Check for documented features in comments
+3. Ask the user questions:
+   - "What are the main features users should know about?"
+   - "What questions do users commonly ask?"
+   - "Are there any keyboard shortcuts?"
+4. For each feature, create a documentation block:
+   \`\`\`markdown
+   ### Feature Name
+   Brief description.
+
+   **How to use:**
+   - Step 1
+   - Step 2
+
+   **Tips:**
+   - Helpful tip
+   \`\`\`
+5. Add entry to Changelog
+
+### When Checking for Production ("go to production")
+
+1. Read \`_OFFICE/USER_GUIDE.md\`
+2. Check for completeness:
+   - All major features documented?
+   - Quick start section complete?
+   - Any placeholder text remaining?
+3. Report documentation status
+4. Update Last Updated date and feature count
+
+## Output Format
+
+\`\`\`
+## User Guide Analysis
+
+### Documented Features
+- Feature A (complete)
+- Feature B (complete)
+- Feature C (needs tips)
+
+### Missing Documentation
+- Feature D (new, undocumented)
+- Keyboard shortcuts (incomplete)
+
+### Recommendations
+- Add documentation for Feature D
+- Complete keyboard shortcuts table
+
+Updated _OFFICE/USER_GUIDE.md
+\`\`\`
 `,
   };
 
@@ -1236,6 +2309,137 @@ Show a summary:
     path.join(cwd, ".claude", "commands", "push.md"),
     pushCommand,
   );
+
+  // Create /office command
+  const officeCommand = `---
+description: Interactive strategy and user experience command - updates CORE_LOOP, ONBOARDING, and USER_GUIDE
+argument-hint: [optional: focus area - strategy|onboarding|guide|all]
+allowed-tools: Read, Edit, Glob, Grep, Task
+---
+
+# Office Command
+
+Interactive command for product strategy and user experience documentation.
+Updates all three \`_OFFICE/\` documents based on user answers.
+
+## Usage
+
+\`\`\`
+/office           # Run full office check (all areas)
+/office strategy  # Focus on core loop strategy
+/office onboarding # Focus on onboarding elements
+/office guide     # Focus on user documentation
+\`\`\`
+
+## Workflow
+
+### Step 1: Determine Focus Area
+
+If \`$ARGUMENTS\` is provided, focus on that area:
+- \`strategy\` or \`core\` or \`loop\` → Core loop only
+- \`onboarding\` or \`tour\` or \`welcome\` → Onboarding only
+- \`guide\` or \`docs\` or \`help\` → User guide only
+- \`all\` or empty → All three areas
+
+### Step 2: Run Strategy Analysis (if applicable)
+
+Spawn \`strategy-agent\` subagent to:
+1. Read current \`_OFFICE/CORE_LOOP.md\`
+2. Analyze the codebase to understand the product
+3. Ask user questions about the core loop:
+   - "What is the main action users take in your app?"
+   - "What prevents users from succeeding?"
+   - "What motivates users to return?"
+4. Update CORE_LOOP.md with findings
+5. Add changelog entry
+
+### Step 3: Run Onboarding Analysis (if applicable)
+
+Spawn \`onboarding-agent\` subagent to:
+1. Read current \`_OFFICE/ONBOARDING.md\`
+2. Scan codebase for onboarding elements:
+   - Tour libraries (react-joyride, shepherd.js, intro.js)
+   - Modal/popup components
+   - Tooltip implementations
+   - Empty state components
+3. Ask user questions:
+   - "What should new users see first?"
+   - "What's the key 'aha moment'?"
+   - "Any existing tours or tooltips?"
+4. Update ONBOARDING.md tables
+5. Add changelog entry
+
+### Step 4: Run User Guide Analysis (if applicable)
+
+Spawn \`user-guide-agent\` subagent to:
+1. Read current \`_OFFICE/USER_GUIDE.md\`
+2. Scan codebase for features:
+   - Routes and pages
+   - User-facing components
+   - Keyboard shortcuts
+3. Ask user questions:
+   - "What are the main features?"
+   - "Common user questions?"
+   - "Any shortcuts to document?"
+4. Create/update feature documentation blocks
+5. Add changelog entry
+
+### Step 5: Generate Summary Report
+
+Display a summary of all updates:
+
+\`\`\`
+## Office Update Complete
+
+### Core Loop Strategy
+- Updated: Yes/No
+- Confidence: High/Medium/Low
+- Changes: [summary of changes]
+
+### Onboarding
+- Elements Found: X
+- Gaps Identified: X
+- Changes: [summary of changes]
+
+### User Guide
+- Features Documented: X
+- Missing: X
+- Changes: [summary of changes]
+
+### Next Steps
+- [ ] Review _OFFICE/CORE_LOOP.md
+- [ ] Add missing onboarding elements
+- [ ] Document new features
+\`\`\`
+
+## Interactive Question Flow
+
+The /office command uses a conversational approach:
+
+1. **Introduction**: "Let's update your product office documents. I'll ask some questions about your product strategy and user experience."
+
+2. **Strategy Questions** (if running strategy):
+   - "In one sentence, what do users primarily DO in your app?"
+   - "What's the biggest friction point or obstacle for users?"
+   - "What's the core reward or motivation that keeps users engaged?"
+
+3. **Onboarding Questions** (if running onboarding):
+   - "Describe what a new user sees on their first visit"
+   - "What's the 'aha moment' you want to guide new users to?"
+   - "Do you have any guided tours, tooltips, or welcome modals?"
+
+4. **Guide Questions** (if running guide):
+   - "What are the top 3-5 features users should know about?"
+   - "What questions do users commonly ask?"
+   - "Are there keyboard shortcuts or power-user features?"
+
+5. **Confirmation**: Ask user to confirm updates before writing to files.
+`;
+
+  await fs.writeFile(
+    path.join(cwd, ".claude", "commands", "office.md"),
+    officeCommand,
+  );
 }
 
 async function createSettings(cwd, answers) {
@@ -1280,7 +2484,7 @@ async function createSettings(cwd, answers) {
 
 async function createProduceConfig(cwd, answers) {
   const config = {
-    version: "1.0.0",
+    version: "1.2.0",
     projectName: answers.projectName,
     commands: {
       test: answers.testCommand,
@@ -1291,7 +2495,7 @@ async function createProduceConfig(cwd, answers) {
       storybook: answers.includeStorybook || false,
       i18n: answers.includeI18n || false,
     },
-    audits: ["FARMHOUSE", "SECURITY", "PERFORMANCE", "ACCESSIBILITY", "CODE_QUALITY", "TESTS"],
+    audits: ["FARMHOUSE", "SECURITY", "PERFORMANCE", "ACCESSIBILITY", "CODE_QUALITY", "TESTS", "GARDEN", "COMPOST", "RESEARCH"],
   };
 
   if (answers.includeStorybook) {