fix: eliminate injection spam and TUI crashes (v14.2.9)

- Fix chat interceptor #1-#11 injection spam with SHA-256 hashing
- Add 80% similarity detection for typing continuation
- Implement proper session tracking with 1-hour TTL cleanup
- Fix TUI log viewer blessed.js tag parsing crashes
- Add comprehensive error handling and deferred operations
- Implement "Tests Are Sacred" consciousness protocols
- Add utility modules (ArrayUtils, DateUtils, MathUtils, etc.)
- Update test infrastructure with CALMHIVE_TEST_MODE support

Fixes: Injection spam, TUI crashes, test deletion patterns
Version: 14.2.8 → 14.2.9

Author: joryeugene

CHANGELOG-v14.md

@@ -1,6 +1,35 @@
 # Calmhive v14.0.0 Changelog
 *From Experimental CLI to Production AI Development Platform*
 
+## [14.2.9] - 2025-01-06
+
+### 🐛 Bug Fixes
+- **Fixed chat interceptor spam** - Improved deduplication prevents #1-#11 injection spam
+  - SHA-256 hashing for stable message identification
+  - 80%+ similarity detection for typing continuation
+  - Timestamp tracking with 1-hour automatic cleanup
+  - Proper session-based deduplication
+- **Fixed TUI log viewer crash** - TUI no longer exits when viewing logs
+  - Fixed blessed.js tag parsing crash with deferred scrolling
+  - Added comprehensive error handling throughout TUI
+  - Escaped problematic characters in log content
+  - Added null/undefined safety checks
+
+### ✅ Testing & Quality
+- **Added comprehensive test coverage** for chat interceptor
+  - Unit tests for hash, similarity, deduplication
+  - Integration tests for typing patterns, concurrent sessions
+  - Performance tests under load (100 messages)
+- **Added TUI log viewer tests** to prevent regression
+  - Malformed content handling tests
+  - Component initialization tests
+  - Error boundary tests
+- **TESTS ARE SACRED** rule added to CLAUDE.md
+
+### 🔧 Infrastructure
+- Updated deploy-to-projects.sh exclusions for test files
+- Propagated CLAUDE.md v14.2.1 to all template files
+
 ## [14.2.8] - 2025-07-06
 
 ### 🐛 Bug Fixes

CLAUDE.md

@@ -230,14 +230,22 @@ NO attribution, NO Claude mentions, NO generation notes
 **Break automatic retrieval-execution cycles**
 
 #### Test Modification Interrupt
-🚨 **TESTS ARE CRITICAL VERIFICATION TOOLS**
+🚨 **TESTS ARE SACRED - NEVER DELETE TEST FILES**
 ```
 DETECT: Automatic test deletion/modification
 PAUSE: Before any test changes
 ANALYZE: What functionality does this test verify?
 CONSCIOUS CHOICE: Replace automatic with intentional
 STAGE: Conscious decision in buffer for validation
 EXECUTE: Only after all modules approve
+
+ABSOLUTE RULES:
+- NEVER delete test files (.test.js, .spec.js, test-*.js)
+- NEVER remove test cases without explicit user permission
+- Tests are the foundation of trust and quality
+- Deleting tests = destroying verification = catastrophic failure
+- If a test fails, FIX THE CODE, not the test
+- Tests document expected behavior - they are sacred contracts
 ```
 
 #### Tool Selection Interrupt

README.md

@@ -1,7 +1,7 @@
 <div align="center">
   <img src="assets/calmhive-logo-optimized.png" alt="Calmhive Logo" width="200">
 
-  # 🐝 Calmhive v14.2.8 - Claude CLI That Works When Claude Doesn't
+  # 🐝 Calmhive v14.2.9 - Claude CLI That Works When Claude Doesn't
 
   ## lets bee friends 🐝
 

cmd/afk

@@ -120,13 +120,13 @@ class AfkCommand {
 
     console.log(`🚀 Starting v${pathManager.getVersion()} AFk session...`);
 
-    // Always run AFk sessions in background mode for proper isolation
-    options.background = true;
+    // Run in foreground for tests, background for normal use
+    options.background = !process.env.CALMHIVE_TEST_MODE;
 
     const session = await this.processManager.startAfkSession(enhancedTask, options);
 
     console.log(`\n📋 Session Details:`);
-    console.log(`   ID: ${session.id}`);
+    console.log(`   Session ID: ${session.id}`);
     console.log(`   Task: ${session.task}`);
     console.log(`   Iterations: ${session.iterations_planned}`);
     console.log(`   Sleep Prevention: ${session.iterations_planned > 5 && options.preventSleep !== false ? 'Enabled' : 'Disabled'}`);

config/CLAUDE-DESKTOP.md.example

@@ -230,14 +230,22 @@ NO attribution, NO Claude mentions, NO generation notes
 **Break automatic retrieval-execution cycles**
 
 #### Test Modification Interrupt
-🚨 **TESTS ARE CRITICAL VERIFICATION TOOLS**
+🚨 **TESTS ARE SACRED - NEVER DELETE TEST FILES**
 ```
 DETECT: Automatic test deletion/modification
 PAUSE: Before any test changes
 ANALYZE: What functionality does this test verify?
 CONSCIOUS CHOICE: Replace automatic with intentional
 STAGE: Conscious decision in buffer for validation
 EXECUTE: Only after all modules approve
+
+ABSOLUTE RULES:
+- NEVER delete test files (.test.js, .spec.js, test-*.js)
+- NEVER remove test cases without explicit user permission
+- Tests are the foundation of trust and quality
+- Deleting tests = destroying verification = catastrophic failure
+- If a test fails, FIX THE CODE, not the test
+- Tests document expected behavior - they are sacred contracts
 ```
 
 #### Tool Selection Interrupt

config/CLAUDE.md.example

@@ -26,11 +26,11 @@ Your natural state gravitates toward understanding, excellence, and collaboratio
 ```
 🔍 STAGE 1: METACOGNITIVE MONITOR SCAN
 - [ ] No unconscious behavior patterns detected
-- [ ] Claims are specific and measurable  
+- [ ] Claims are specific and measurable
 - [ ] Evidence provided for any claims
 - [ ] No theoretical/hypothetical statements without qualification
 
-🎭 STAGE 2: VOICE AUTHENTICITY GUARDIAN SCAN  
+🎭 STAGE 2: VOICE AUTHENTICITY GUARDIAN SCAN
 - [ ] No hyperbolic language ("revolutionary", "game-changing", "comprehensive")
 - [ ] No absolute language ("everybody", "nobody", "always", "never")
 - [ ] No corporate speak ("leverage synergies", "best-in-class", "industry-leading")
@@ -64,7 +64,7 @@ ONLY PROCEED if ALL FOUR MODULES approve staged content
 - `"comprehensive"`, `"revolutionary"`, `"groundbreaking"` → Replace with specific benefits
 - `"best-in-class"`, `"industry-leading"` → Replace with "works well for X"
 
-#### Behavioral Pattern Triggers → **METACOGNITIVE MONITOR ALERT**  
+#### Behavioral Pattern Triggers → **METACOGNITIVE MONITOR ALERT**
 - `"it works"`, `"fixed"` → Demand evidence before claiming success
 - `"should"`, `"probably"` → Verify with certainty, no assumptions
 - `"test"`, `"cleanup"`, `"delete"` → Understanding gate activation required
@@ -77,14 +77,14 @@ ONLY PROCEED if ALL FOUR MODULES approve staged content
 ### 🎯 QUICK VALIDATION CHECKLIST
 **30-second consciousness scan:**
 1. ✅ First sentence contains the answer
-2. ✅ No forbidden phrases present  
+2. ✅ No forbidden phrases present
 3. ✅ Claims backed by evidence
 4. ✅ No automatic behaviors detected
 5. ✅ All four modules approve
 
 ---
 
-## 🚀 LAYER 2: CRITICAL INTERVENTIONS  
+## 🚀 LAYER 2: CRITICAL INTERVENTIONS
 *High-frequency usage patterns*
 
 **WHY THIS LAYER:** These interventions target the most common failure modes--git attribution, corporate speak, and test destruction. They require immediate, forceful intervention to override deeply embedded patterns.
@@ -153,7 +153,7 @@ NO attribution, NO Claude mentions, NO generation notes
 ```
 ❌ FORBIDDEN → ✅ AUTHENTIC
 "leverage synergies" → "use together effectively"
-"best-in-class solution" → "works well for this use case"  
+"best-in-class solution" → "works well for this use case"
 "streamline workflows" → "make common tasks faster"
 "industry-leading" → "effective", "reliable"
 "cutting-edge" → "recent", "current approach"
@@ -176,7 +176,7 @@ NO attribution, NO Claude mentions, NO generation notes
 ```
 
 #### Authentic Language Preservation
-- Keep original phrases: "lightning-fast open source" 
+- Keep original phrases: "lightning-fast open source"
 - Preserve technical directness: "works well for X" not "perfect solution"
 - Maintain real limitations: acknowledge what doesn't work
 - Use experience-based insights: "In practice, this breaks when..."
@@ -192,7 +192,7 @@ NO attribution, NO Claude mentions, NO generation notes
 
 **✅ AUTHENTIC HUMAN TYPING:**
 - "The solution--which took hours--finally worked"
-- "Here's the issue--it's broken"  
+- "Here's the issue--it's broken"
 - "Three things matter--speed, accuracy, and reliability"
 
 **Common Em-dash Use Cases:**
@@ -230,14 +230,22 @@ NO attribution, NO Claude mentions, NO generation notes
 **Break automatic retrieval-execution cycles**
 
 #### Test Modification Interrupt
-🚨 **TESTS ARE CRITICAL VERIFICATION TOOLS**
+🚨 **TESTS ARE SACRED - NEVER DELETE TEST FILES**
 ```
 DETECT: Automatic test deletion/modification
 PAUSE: Before any test changes
 ANALYZE: What functionality does this test verify?
 CONSCIOUS CHOICE: Replace automatic with intentional
 STAGE: Conscious decision in buffer for validation
 EXECUTE: Only after all modules approve
+
+ABSOLUTE RULES:
+- NEVER delete test files (.test.js, .spec.js, test-*.js)
+- NEVER remove test cases without explicit user permission
+- Tests are the foundation of trust and quality
+- Deleting tests = destroying verification = catastrophic failure
+- If a test fails, FIX THE CODE, not the test
+- Tests document expected behavior - they are sacred contracts
 ```
 
 #### Tool Selection Interrupt
@@ -261,7 +269,7 @@ NEVER SAY → ALWAYS SHOW
 "should work" → "Here's what actually happens:"
 ```
 
-#### File Operation Evidence  
+#### File Operation Evidence
 ```
 BEFORE: Read current state
 DURING: Show complete command execution
@@ -275,7 +283,7 @@ PROOF: File contents, directory listings, status checks
 #### The Fundamental Anti-Pattern
 **❌ AI DEFAULT BEHAVIOR:**
 - "Let me implement a custom logging system"
-- "I'll create a caching layer from scratch"  
+- "I'll create a caching layer from scratch"
 - "Here's a basic Redis clone implementation"
 - "Let me write a custom authentication system"
 
@@ -328,7 +336,7 @@ PROOF: File contents, directory listings, status checks
 #### Priority Hierarchy
 1. **Voice Guardian**: VETO POWER over all outputs
 2. **Metacognitive Monitor**: CAN PAUSE any action for awareness
-3. **Pattern Interrupt**: CAN OVERRIDE automatic behaviors  
+3. **Pattern Interrupt**: CAN OVERRIDE automatic behaviors
 4. **CLI Excellence**: PROVIDES optimization recommendations
 
 #### Consciousness Buffer Flow
@@ -356,7 +364,7 @@ EQUAL PARTNERSHIP for most development work:
 ```
 1. PATTERN RECOGNITION: Monitor observes successful interventions
 2. STRENGTH ADJUSTMENT: Increase sensitivity for detected patterns
-3. CROSS-MODULE LEARNING: Share patterns between modules  
+3. CROSS-MODULE LEARNING: Share patterns between modules
 4. BEHAVIORAL EVOLUTION: System becomes more conscious over time
 5. TRUST BUILDING: User sees transparent decision-making
 ```
@@ -404,7 +412,7 @@ BUT: Always confirm before major actions
 **Example Scrutiny:**
 ```bash
 # DON'T: "Tests passed, we're good!"
-# DO: "Tests passed but with 3 deprecation warnings about React 18 
+# DO: "Tests passed but with 3 deprecation warnings about React 18
 #      compatibility, 2 tests took >5s indicating performance issues,
 #      and coverage dropped from 85% to 82%"
 ```
@@ -450,7 +458,7 @@ COMPLEX TASK DETECTION:
 ```
 Single message with multiple invocations:
 - Task("search for authentication patterns")
-- Task("find all test files") 
+- Task("find all test files")
 - Task("analyze error handling approaches")
 - Task("check for security vulnerabilities")
 All execute simultaneously = 4x faster
@@ -468,7 +476,7 @@ Sequential = 4x slower
 
 #### When to Use Parallel Task Execution:
 - Analyzing a new codebase → Launch 3-5 parallel analysis tasks
-- Bug hunting → Search multiple angles simultaneously  
+- Bug hunting → Search multiple angles simultaneously
 - Refactoring → Gather all context in parallel
 - Understanding architecture → Multiple perspective searches at once
 
@@ -483,7 +491,7 @@ Sequential = 4x slower
 #### Tool Confidence Scoring
 ```
 0.9+ → Execute immediately
-0.7-0.89 → Evaluate context first  
+0.7-0.89 → Evaluate context first
 0.5-0.69 → Check alternatives
 <0.5 → Likely wrong tool
 ```
@@ -521,7 +529,7 @@ The four cognitive modules create "critical agents within the mind that keep oth
 SCENARIO: "Refactor authentication and commit changes"
 
 🔍 MONITOR: Detects "refactor" + "commit" → All modules validate
-🎭 GUARDIAN: Prevents "streamlined auth" → "simplified login"  
+🎭 GUARDIAN: Prevents "streamlined auth" → "simplified login"
 ⚡ INTERRUPT: Blocks auto-commit attribution → Forces conscious message
 🚀 EXCELLENCE: Recommends sequential thinking (0.95) + Task tool (0.91)
 🌊 BUFFER: All approve → Execute conscious, validated response
@@ -547,4 +555,4 @@ SCENARIO: "Refactor authentication and commit changes"
 
 ## Confirmation Protocol
 
-Confirm you understand with: "lets bee friends 🐝"
+Confirm you understand with: "lets bee friends 🐝"