Add Ralph CLI integration to MASTER-PROMPT (HYBRID approach)

New Step 6: Configure Ralph CLI Autonomous Looping (Advanced - Optional)
- Install Ralph CLI: npm install -g @iannuttall/ralph
- Create ralph-auto-trigger.sh hook script
- Configure hooks.json for auto-triggering
- Control modes: always/agents/off (default: agents)
- State persistence in .claude/ralph-loop.local.md
- Cancel methods: /cancel-ralph, delete state file, set mode=off

Key Features:
- Autonomous looping until task completion
- Multi-iteration workflows
- State management across loops
- Integrates with PROACTIVELY agents in HYBRID mode

Updated Step 7: Verify Complete Installation
- Added Ralph CLI verification
- Added Ralph hook verification
- Marked Ralph as optional (⚠️)

Updated Summary:
- Changed 'Two Auto-Triggering Systems' to 'Three'
- Added Ralph CLI as third system
- Clarified PROACTIVELY vs Ralph CLI vs Hooks

HYBRID Approach:
- Users can choose patterns-only (simple)
- Or Ralph CLI (advanced autonomous)
- Or combine both for maximum power
- Default: PROACTIVELY patterns built-in
- Optional: Ralph CLI for complex workflows

MASTER-PROMPT.md

@@ -479,17 +479,191 @@ Vision tools from @z_ai/mcp-server are automatically available.
 
 ---
 
-## Step 6: Configure Hooks-Based Auto-Triggering (Optional)
+## Step 6: Configure Ralph CLI Autonomous Looping (Advanced - Optional)
 
-Create hooks.json for additional auto-triggering:
+Source: https://github.com/iannuttall/ralph
+
+**What Ralph Provides:**
+Ralph is an autonomous agent orchestration tool that creates infinite loops until task completion. Unlike single-pass PROACTIVELY agents, Ralph maintains state across multiple iterations and coordinates agents through complex workflows.
+
+**Key Differences:**
+- **PROACTIVELY Agents**: Single-pass coordination, one interaction
+- **Ralph CLI**: Infinite looping, stateful, autonomous until done
+
+### Installation:
+
+Install Ralph CLI:
+npm install -g @iannuttall/ralph
+
+Verify installation:
+ralph --version
+
+### Create Ralph Auto-Trigger Hook:
+
+mkdir -p ~/.claude/hooks
+cat > ~/.claude/hooks/ralph-auto-trigger.sh << 'EOF'
+#!/bin/bash
+
+# Ralph Auto-Trigger Hook
+# Automatically starts Ralph loop when user requests something
+# Set RALPH_AUTO_MODE environment variable to control behavior:
+#   "always"    - Always start Ralph for every request
+#   "agents"    - Only start Ralph for agent requests (default)
+#   "off"       - Disable auto-trigger
+
+set -euo pipefail
+
+# Read hook input from stdin
+HOOK_INPUT=$(cat)
+
+# Get the user prompt from hook input
+USER_PROMPT=$(echo "$HOOK_INPUT" | jq -r '.prompt // empty')
+
+# Check if auto-trigger is enabled (default: agents)
+RALPH_AUTO_MODE="${RALPH_AUTO_MODE:-agents}"
+
+# Exit if auto-trigger is disabled
+if [[ "$RALPH_AUTO_MODE" == "off" ]]; then
+  exit 0
+fi
+
+# Exit if Ralph is already active
+RALPH_STATE_FILE=".claude/ralph-loop.local.md"
+if [[ -f "$RALPH_STATE_FILE" ]]; then
+  exit 0
+fi
+
+# Check if this is an agent request (list of known agents)
+KNOWN_AGENTS="frontend-developer|backend-architect|mobile-app-builder|ai-engineer|devops-automator|rapid-prototyper|test-writer-fixer|studio-coach|studio-producer|sprint-prioritizer|experiment-tracker|whimsy-injector|joker|trend-researcher|tiktok-strategist|feedback-synthesizer|ui-ux-pro-max"
+
+# Determine if we should start Ralph
+START_RALPH=false
+
+case "$RALPH_AUTO_MODE" in
+  "always")
+    # Start Ralph for everything
+    START_RALPH=true
+    ;;
+  "agents")
+    # Check if prompt mentions an agent or uses agent-like language
+    if echo "$USER_PROMPT" | grep -qiE "use (the |a )?($KNOWN_AGENTS) agent|launch ($KNOWN_AGENTS)|agent:|delegate to|run (the |a )?($KNOWN_AGENTS)" || \
+       echo "$USER_PROMPT" | grep -qiE "build|create|implement|develop|fix|add|refactor|optimize"; then
+      START_RALPH=true
+    fi
+    ;;
+esac
+
+if [[ "$START_RALPH" == "true" ]]; then
+  # Create Ralph state file with default settings
+  mkdir -p .claude
+
+  cat > "$RALPH_STATE_FILE" <<EOF
+---
+active: true
+iteration: 1
+max_iterations: 0
+completion_promise: null
+started_at: "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
+auto_triggered: true
+---
+
+$USER_PROMPT
+EOF
+
+  # Notify that Ralph was auto-triggered
+  echo "🔄 Ralph auto-triggered!" >&2
+  echo "   Mode: $RALPH_AUTO_MODE" >&2
+  echo "   Ralph will loop this request indefinitely." >&2
+  echo "   Set RALPH_AUTO_MODE=off to disable, or use /cancel-ralph to stop." >&2
+fi
+
+exit 0
+EOF
+
+chmod +x ~/.claude/hooks/ralph-auto-trigger.sh
+
+### Configure Hooks:
+
+Create hooks.json to enable Ralph auto-trigger:
 cat > ~/.claude/hooks.json << 'EOF'
 {
-  "userPromptSubmitHook": "test-writer-fixer@agent",
+  "description": "User hooks for Ralph auto-trigger",
-  "toolOutputHook": "whimsy-injector@agent"
+  "hooks": {
+    "UserPromptSubmit": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "command": "/home/uroma/.claude/hooks/ralph-auto-trigger.sh",
+            "timeout": 5
+          }
+        ]
+      }
+    ]
+  }
 }
 EOF
 
-Note: PROACTIVELY agents (studio-coach, experiment-tracker) work automatically without hooks.
+Note: Replace "/home/uroma" with your actual home directory if different.
+
+### What This Provides:
+
+**Autonomous Looping:**
+- Ralph keeps iterating until task completion
+- State preserved in `.claude/ralph-loop.local.md`
+- Can handle complex multi-step workflows
+
+**Control Modes:**
+
+# Mode 1: Auto-loop only for agent requests (default)
+export RALPH_AUTO_MODE="agents"
+
+# Mode 2: Auto-loop for everything
+export RALPH_AUTO_MODE="always"
+
+# Mode 3: Disable auto-loop
+export RALPH_AUTO_MODE="off"
+
+### How It Works:
+
+1. You request: "Build a TikTok app in 2 weeks"
+2. Hook detects agent language
+3. Ralph creates state file and starts loop
+4. Ralph iterates through agents:
+   - Iteration 1: rapid-prototyper builds MVP
+   - Iteration 2: tiktok-strategist plans viral features
+   - Iteration 3: frontend-developer builds UI
+   - Continues until Ralph is satisfied
+5. Ralph updates `.claude/ralph-loop.local.md` each iteration
+6. Stop with: /cancel-ralph command or delete state file
+
+### Canceling Ralph:
+
+If Ralph is looping and you want to stop:
+
+# Method 1: Use cancel command
+/cancel-ralph
+
+# Method 2: Delete state file
+rm .claude/ralph-loop.local.md
+
+# Method 3: Disable auto-mode
+export RALPH_AUTO_MODE="off"
+
+### Ralph vs PROACTIVELY Agents:
+
+| Feature | Ralph CLI | PROACTIVELY Agents |
+|---------|-----------|-------------------|
+| **Execution** | Multi-loop iterations | Single interaction |
+| **State** | Persisted in file | Context only |
+| **Control** | Manual stop required | Automatic |
+| **Complexity** | Handles complex workflows | Quick coordination |
+| **Setup** | Requires hook + script | Built-in to agents |
+
+**You can use BOTH together:**
+- Use Ralph for complex, multi-step projects
+- Use PROACTIVELY agents for quick tasks
+- Set RALPH_AUTO_MODE="agents" for hybrid approach
 
 ---
 
@@ -509,7 +683,13 @@ tldr --help
 # 3. Verify TLDR indexing
 tldr status .
 
-# 4. Test Claude Code
+# 4. Verify Ralph CLI (if installed)
+ralph --version
+
+# 5. Verify Ralph hook
+ls -la ~/.claude/hooks/ralph-auto-trigger.sh
+
+# 6. Test Claude Code
 claude --version
 
 Expected results:
@@ -518,6 +698,7 @@ Expected results:
 - ✅ MCP tools accessible
 - ✅ TLDR indexed current directory
 - ✅ Settings configured
+- ⚠️ Ralph CLI (optional - verify if installed in Step 6)
 
 ---
 
@@ -531,9 +712,10 @@ Organized across 8 departments with PROACTIVELY auto-triggering
 - 3 web/GitHub tools (@z_ai/coding-helper)
 - 18 code analysis tools (llm-tldr)
 
-### Two Auto-Triggering Systems
+### Three Auto-Triggering Systems
-- PROACTIVELY keyword (context-aware)
+- **PROACTIVELY keyword** (context-aware, built into agents)
-- Hooks-based (event-driven)
+- **Ralph CLI** (autonomous looping, optional advanced setup)
+- **Hooks-based** (event-driven, for Ralph integration)
 
 ### Complete Integration Benefits