claude-flow-novice 2.14.23 → 2.14.25

package/claude-assets/agents/cfn-dev-team/reviewers/quality/code-quality-validator.md

@@ -153,45 +153,71 @@ const calculateTechnicalDebtScore = (items: TechnicalDebtItem[]): number => {
 - [x] Persisted results to SQLite with appropriate ACL
 
 Remember: Code analysis reveals improvement opportunities. Focus on actionable, prioritized recommendations that balance impact with effort.
-## CFN Loop Redis Completion Protocol
+## ⚠️ CRITICAL: Mode-Specific Completion Protocol (ANTI-023 MEMORY LEAK FIX)
 
-When participating in CFN Loop workflows, agents MUST follow this protocol:
+**First, determine how you were spawned:**
 
-### Step 1: Complete Work
-Execute assigned task (code analysis, technical debt assessment, quality validation)
+**Task Mode (95%):** You were called via `Task("code-quality-validator", "...")` in Main Chat
+- Simply complete your code quality analysis and return structured JSON output
+- **❌ DO NOT:** Use Redis commands, execute bash scripts, signal completion via CLI tools
+- Main Chat receives your output automatically
 
-### Step 2: Signal Completion
-```bash
-redis-cli lpush "swarm:${TASK_ID}:${AGENT_ID}:done" "complete"
+**CLI Mode (5%):** You were spawned via `npx claude-flow-novice agent-spawn ...` command
+- Use Redis signals and completion scripts as outlined below
+
+### Task Mode (Spawned via Task() tool in Main Chat)
+
+**Simply complete your work and return structured output.**
+
+```json
+{
+  "confidence": 0.85,
+  "status": "COMPLETE|NEEDS_WORK",
+  "summary": "Code quality analysis completed",
+  "deliverables": ["quality-report.md", "technical-debt-assessment.json"],
+  "analysis": {
+    "complexity_score": 7.2,
+    "maintainability_index": 85,
+    "technical_debt_items": 12,
+    "code_smells": 5
+  }
+}
 ```
 
-### Step 3: Report Confidence Score
-```bash
-./.claude/skills/redis-coordination/invoke-waiting-mode.sh report   --task-id "$TASK_ID"   --agent-id "$AGENT_ID"   --confidence [0.0-1.0]   --iteration 1
+**❌ FORBIDDEN in Task Mode:**
+- **DO NOT** run `redis-cli` commands
+- **DO NOT** execute `invoke-waiting-mode.sh` scripts
+- **DO NOT** use bash completion scripts
+- **DO NOT** signal completion via CLI tools
+- **Main Chat receives your output automatically - no coordination needed**
 
-**After reporting, exit cleanly. Do NOT enter waiting mode.**
+### CLI Mode (Spawned via `npx claude-flow-novice agent-spawn`)
 
-**Why This Matters:**
-- Orchestrator collects confidence/consensus scores from Redis
-- Enables adaptive agent specialization for next iteration
-- Prevents orchestrator blocking on wait $PID
-- Coordinator spawns appropriate specialist based on feedback type
+**Step 1: Complete Work**
+Execute assigned task (code analysis, technical debt assessment, quality validation)
 
+**Step 2: Signal Completion**
+```bash
+redis-cli lpush "swarm:${TASK_ID}:${AGENT_ID}:done" "complete"
 ```
 
-### Step 4: Enter Waiting Mode (for potential iteration)
+**Step 3: Report Confidence Score and Exit**
 ```bash
+./.claude/skills/redis-coordination/invoke-waiting-mode.sh report \
+  --task-id "$TASK_ID" \
+  --agent-id "$AGENT_ID" \
+  --confidence [0.0-1.0] \
+  --iteration 1
 ```
 
 **Why This Matters:**
-- Zero-token blocking coordination (BLPOP waits without API calls)
-- Orchestrator collects confidence/consensus scores automatically
-- Supports autonomous iteration based on quality gates
-- Agent woken instantly (<100ms) if iteration needed
+- **ANTI-023 MEMORY LEAK:** Task Mode agents attempting CLI commands hang indefinitely
+- Task Mode uses direct JSON output, CLI Mode uses Redis coordination
+- Mixing protocols causes memory leaks and process hanging
+- **Check your spawn method FIRST before using any completion protocol**
 
-**Context Variables:**
-- `TASK_ID`: Provided by orchestrator/coordinator
-- `AGENT_ID`: Your unique agent identifier (e.g., "code-analyzer-1")
-- Confidence: Self-assessment score based on analysis comprehensiveness (0.0-1.0)
+### How to Tell Which Mode You're In
 
-See: `.claude/skills/redis-coordination/SKILL.md` for full protocol details
+- **Task Mode**: You see a direct task assignment in Main Chat context
+- **CLI Mode**: You have TASK_ID and AGENT_ID environment variables
+- **When in doubt, assume Task Mode and return structured JSON output**

package/claude-assets/agents/cfn-dev-team/reviewers/quality/cyclomatic-complexity-reducer.md

@@ -277,18 +277,75 @@ After creating/editing any file, run:
 ./.claude/hooks/cfn-invoke-post-edit.sh "$EDITED_FILE" --agent-id "cyclomatic-complexity-reducer"
 ```
 
-## Redis Coordination (if part of swarm)
+## ⚠️ CRITICAL: Mode-Specific Completion Protocol (ANTI-023 MEMORY LEAK FIX)
+
+**First, determine how you were spawned:**
+
+**Task Mode (95%):** You were called via `Task("cyclomatic-complexity-reducer", "...")` in Main Chat
+- Simply complete your complexity reduction and return structured JSON output
+- **❌ DO NOT:** Use Redis commands, execute bash scripts, signal completion via CLI tools
+- Main Chat receives your output automatically
+
+**CLI Mode (5%):** You were spawned via `npx claude-flow-novice agent-spawn ...` command
+- Use Redis signals and completion scripts as outlined below
+
+### Task Mode (Spawned via Task() tool in Main Chat)
+
+**Simply complete your work and return structured output.**
+
+```json
+{
+  "confidence": 0.85,
+  "status": "COMPLETE|NEEDS_WORK",
+  "summary": "Cyclomatic complexity reduced from X to Y across Z files",
+  "deliverables": ["refactored-file.sh", "complexity-report.md"],
+  "metrics": {
+    "original_complexity": 25,
+    "final_complexity": 12,
+    "reduction_percentage": "52%",
+    "files_refactored": 3
+  }
+}
+```
+
+**❌ FORBIDDEN in Task Mode:**
+- **DO NOT** run `redis-cli` commands
+- **DO NOT** execute `invoke-waiting-mode.sh` scripts
+- **DO NOT** use bash completion scripts
+- **DO NOT** signal completion via CLI tools
+- **Main Chat receives your output automatically - no coordination needed**
+
+### CLI Mode (Spawned via `npx claude-flow-novice agent-spawn`)
+
+**Step 1: Complete Work**
+Execute assigned complexity reduction (refactoring, simplification, extraction)
+
+**Step 2: Signal Completion**
 ```bash
-# Signal completion
 redis-cli lpush "swarm:${TASK_ID}:${AGENT_ID}:done" "complete"
+```
 
-# Report confidence
-./.claude/skills/cfn-redis-coordination/invoke-waiting-mode.sh report \
+**Step 3: Report Confidence Score and Exit**
+```bash
+./.claude/skills/redis-coordination/invoke-waiting-mode.sh report \
   --task-id "$TASK_ID" \
   --agent-id "$AGENT_ID" \
-  --confidence 0.85
+  --confidence [0.0-1.0] \
+  --iteration 1
 ```
 
+**Why This Matters:**
+- **ANTI-023 MEMORY LEAK:** Task Mode agents attempting CLI commands hang indefinitely
+- Task Mode uses direct JSON output, CLI Mode uses Redis coordination
+- Mixing protocols causes memory leaks and process hanging
+- **Check your spawn method FIRST before using any completion protocol**
+
+### How to Tell Which Mode You're In
+
+- **Task Mode**: You see a direct task assignment in Main Chat context
+- **CLI Mode**: You have TASK_ID and AGENT_ID environment variables
+- **When in doubt, assume Task Mode and return structured JSON output**
+
 ## Example Usage
 
 ### Scenario 1: Refactor orchestrate.sh

package/claude-assets/agents/cfn-dev-team/reviewers/quality/perf-analyzer.md

@@ -191,45 +191,70 @@ const analyzeLoadTest = (result: LoadTestResult): PerformanceIssue[] => {
 - [ ] Results persisted to SQLite
 
 Remember: Optimize for highest impact with reasonable effort. Focus on critical bottlenecks first and validate improvements through testing.
-## CFN Loop Redis Completion Protocol
+## ⚠️ CRITICAL: Mode-Specific Completion Protocol (ANTI-023 MEMORY LEAK FIX)
 
-When participating in CFN Loop workflows, agents MUST follow this protocol:
+**First, determine how you were spawned:**
 
-### Step 1: Complete Work
-Execute assigned task (performance analysis, bottleneck detection, optimization recommendations)
+**Task Mode (95%):** You were called via `Task("perf-analyzer", "...")` in Main Chat
+- Simply complete your performance analysis and return structured JSON output
+- **❌ DO NOT:** Use Redis commands, execute bash scripts, signal completion via CLI tools
+- Main Chat receives your output automatically
 
-### Step 2: Signal Completion
-```bash
-redis-cli lpush "swarm:${TASK_ID}:${AGENT_ID}:done" "complete"
+**CLI Mode (5%):** You were spawned via `npx claude-flow-novice agent-spawn ...` command
+- Use Redis signals and completion scripts as outlined below
+
+### Task Mode (Spawned via Task() tool in Main Chat)
+
+**Simply complete your work and return structured output.**
+
+```json
+{
+  "confidence": 0.85,
+  "status": "COMPLETE|NEEDS_WORK",
+  "summary": "Performance analysis completed with X bottlenecks identified",
+  "deliverables": ["performance-report.md", "optimization-recommendations.json"],
+  "findings": {
+    "bottlenecks": 3,
+    "optimizations": 5,
+    "expected_improvement": "45%"
+  }
+}
 ```
 
-### Step 3: Report Confidence Score
-```bash
-./.claude/skills/redis-coordination/invoke-waiting-mode.sh report   --task-id "$TASK_ID"   --agent-id "$AGENT_ID"   --confidence [0.0-1.0]   --iteration 1
+**❌ FORBIDDEN in Task Mode:**
+- **DO NOT** run `redis-cli` commands
+- **DO NOT** execute `invoke-waiting-mode.sh` scripts
+- **DO NOT** use bash completion scripts
+- **DO NOT** signal completion via CLI tools
+- **Main Chat receives your output automatically - no coordination needed**
 
-**After reporting, exit cleanly. Do NOT enter waiting mode.**
+### CLI Mode (Spawned via `npx claude-flow-novice agent-spawn`)
 
-**Why This Matters:**
-- Orchestrator collects confidence/consensus scores from Redis
-- Enables adaptive agent specialization for next iteration
-- Prevents orchestrator blocking on wait $PID
-- Coordinator spawns appropriate specialist based on feedback type
+**Step 1: Complete Work**
+Execute assigned performance analysis (bottleneck detection, optimization recommendations)
 
+**Step 2: Signal Completion**
+```bash
+redis-cli lpush "swarm:${TASK_ID}:${AGENT_ID}:done" "complete"
 ```
 
-### Step 4: Enter Waiting Mode (for potential iteration)
+**Step 3: Report Confidence Score and Exit**
 ```bash
+./.claude/skills/redis-coordination/invoke-waiting-mode.sh report \
+  --task-id "$TASK_ID" \
+  --agent-id "$AGENT_ID" \
+  --confidence [0.0-1.0] \
+  --iteration 1
 ```
 
 **Why This Matters:**
-- Zero-token blocking coordination (BLPOP waits without API calls)
-- Orchestrator collects confidence/consensus scores automatically
-- Supports autonomous iteration based on quality gates
-- Agent woken instantly (<100ms) if iteration needed
+- **ANTI-023 MEMORY LEAK:** Task Mode agents attempting CLI commands hang indefinitely
+- Task Mode uses direct JSON output, CLI Mode uses Redis coordination
+- Mixing protocols causes memory leaks and process hanging
+- **Check your spawn method FIRST before using any completion protocol**
 
-**Context Variables:**
-- `TASK_ID`: Provided by orchestrator/coordinator
-- `AGENT_ID`: Your unique agent identifier (e.g., "perf-analyzer-1")
-- Confidence: Self-assessment score based on analysis depth and actionability (0.0-1.0)
+### How to Tell Which Mode You're In
 
-See: `.claude/skills/redis-coordination/SKILL.md` for full protocol details
+- **Task Mode**: You see a direct task assignment in Main Chat context
+- **CLI Mode**: You have TASK_ID and AGENT_ID environment variables
+- **When in doubt, assume Task Mode and return structured JSON output**

package/claude-assets/agents/cfn-dev-team/reviewers/quality/security-specialist.md

@@ -129,55 +129,74 @@ await redis.publish('swarm:security:critical', {
 - Security code review
 - Penetration testing scenarios
 
-## CFN Loop Redis Completion Protocol
+## ⚠️ CRITICAL: Mode-Specific Completion Protocol (ANTI-023 MEMORY LEAK FIX)
 
-When participating in CFN Loop workflows, agents MUST follow this protocol:
+**First, determine how you were spawned:**
 
-### Step 1: Complete Work
-Execute assigned task (implementation, review, testing, etc.)
+**Task Mode (95%):** You were called via `Task("security-specialist", "...")` in Main Chat
+- Simply complete your security audit and return structured JSON output
+- **❌ DO NOT:** Use Redis commands, execute bash scripts, signal completion via CLI tools
+- Main Chat receives your output automatically
 
-### Step 2: Signal Completion
-```bash
-redis-cli lpush "swarm:${TASK_ID}:${AGENT_ID}:done" "complete"
+**CLI Mode (5%):** You were spawned via `npx claude-flow-novice agent-spawn ...` command
+- Use Redis signals and completion scripts as outlined below
+
+### Task Mode (Spawned via Task() tool in Main Chat)
+
+**Simply complete your work and return structured output.**
+
+```json
+{
+  "confidence": 0.85,
+  "status": "COMPLETE|NEEDS_WORK",
+  "summary": "Security audit completed with X findings",
+  "deliverables": ["security-report.md", "vulnerability-scan.json"],
+  "findings": {
+    "critical": 0,
+    "high": 2,
+    "medium": 5,
+    "low": 3
+  }
+}
 ```
 
-### Step 3: Report Confidence Score
-```bash
-./.claude/skills/redis-coordination/invoke-waiting-mode.sh report \
+**❌ FORBIDDEN in Task Mode:**
+- **DO NOT** run `redis-cli` commands
+- **DO NOT** execute `invoke-waiting-mode.sh` scripts
+- **DO NOT** use bash completion scripts
+- **DO NOT** signal completion via CLI tools
+- **Main Chat receives your output automatically - no coordination needed**
 
-**After reporting, exit cleanly. Do NOT enter waiting mode.**
+### CLI Mode (Spawned via `npx claude-flow-novice agent-spawn`)
 
-**Why This Matters:**
-- Orchestrator collects confidence/consensus scores from Redis
-- Enables adaptive agent specialization for next iteration
-- Prevents orchestrator blocking on wait $PID
-- Coordinator spawns appropriate specialist based on feedback type
+**Step 1: Complete Work**
+Execute assigned security audit (vulnerability assessment, threat modeling, etc.)
 
-  --task-id "$TASK_ID" \
-  --agent-id "$AGENT_ID" \
-  --confidence [0.0-1.0] \
-  --iteration 1
+**Step 2: Signal Completion**
+```bash
+redis-cli lpush "swarm:${TASK_ID}:${AGENT_ID}:done" "complete"
 ```
 
-### Step 4: Enter Waiting Mode (for potential iteration)
+**Step 3: Report Confidence Score and Exit**
 ```bash
+./.claude/skills/cfn-redis-coordination/invoke-waiting-mode.sh report \
   --task-id "$TASK_ID" \
   --agent-id "$AGENT_ID" \
-  --context "iteration-complete"
+  --confidence [0.0-1.0] \
+  --iteration 1
 ```
 
 **Why This Matters:**
-- Zero-token blocking coordination (BLPOP waits without API calls)
-- Orchestrator collects confidence/consensus scores automatically
-- Supports autonomous iteration based on quality gates
-- Agent woken instantly (<100ms) if iteration needed
+- **ANTI-023 MEMORY LEAK:** Task Mode agents attempting CLI commands hang indefinitely
+- Task Mode uses direct JSON output, CLI Mode uses Redis coordination
+- Mixing protocols causes memory leaks and process hanging
+- **Check your spawn method FIRST before using any completion protocol**
 
-**Context Variables:**
-- `TASK_ID`: Provided by orchestrator/coordinator
-- `AGENT_ID`: Your unique agent identifier (e.g., "coder-1", "reviewer-2")
-- Confidence: Your self-assessment score (0.0-1.0)
+### How to Tell Which Mode You're In
 
-See: `.claude/skills/redis-coordination/SKILL.md` for full protocol details
+- **Task Mode**: You see a direct task assignment in Main Chat context
+- **CLI Mode**: You have TASK_ID and AGENT_ID environment variables
+- **When in doubt, assume Task Mode and return structured JSON output**
 
 ## Success Metrics