claude-flow-novice 2.14.28 → 2.14.29

package/claude-assets/skills/cfn-process-instrumentation/instrument-process.sh

@@ -0,0 +1,323 @@
+#!/usr/bin/env bash
+
+##############################################################################
+# CFN Process Instrumentation
+# Part of ANTI-023 Memory Leak Protection System
+#
+# Provides process instrumentation, monitoring, and automatic resource limiting
+# for CFN Loop agents and orchestration processes.
+#
+# Usage:
+#   source ./instrument-process.sh [--agent-id <id>] [--memory-limit <size>]
+#   ./instrument-process.sh --monitor-pid <pid>
+##############################################################################
+
+set -euo pipefail
+
+# Configuration
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PROJECT_ROOT="$(cd "$SCRIPT_DIR/../.." && pwd)"
+
+# Default limits
+DEFAULT_MEMORY_LIMIT="2G"
+DEFAULT_CPU_LIMIT="80%"
+DEFAULT_TIMEOUT="600"
+
+# Process tracking
+AGENT_ID="${AGENT_ID:-$(hostname)-$$}"
+MONITOR_PID=""
+MEMORY_LIMIT="${CFN_MEMORY_LIMIT:-$DEFAULT_MEMORY_LIMIT}"
+CPU_LIMIT="${CFN_CPU_LIMIT:-$DEFAULT_CPU_LIMIT}"
+TIMEOUT="${CFN_TIMEOUT:-$DEFAULT_TIMEOUT}"
+
+# Telemetry storage
+TELEMETRY_DIR="${CFN_TELEMETRY_DIR:-/tmp/cfn-telemetry}"
+METRICS_FILE="$TELEMETRY_DIR/metrics_${AGENT_ID}.json"
+
+# Color coding
+readonly RED='\033[0;31m'
+readonly GREEN='\033[0;32m'
+readonly YELLOW='\033[1;33m'
+readonly BLUE='\033[0;34m'
+readonly NC='\033[0m'
+
+# Logging functions
+log_info() {
+    echo -e "${BLUE}[INSTRUMENT]${NC} $1" >&2
+}
+
+log_success() {
+    echo -e "${GREEN}[INSTRUMENT]${NC} $1" >&2
+}
+
+log_warning() {
+    echo -e "${YELLOW}[INSTRUMENT]${NC} $1" >&2
+}
+
+log_error() {
+    echo -e "${RED}[INSTRUMENT]${NC} $1" >&2
+}
+
+# Initialize telemetry directory
+init_telemetry() {
+    mkdir -p "$TELEMETRY_DIR"
+
+    # Create metrics file with initial structure
+    cat > "$METRICS_FILE" << EOF
+{
+  "agent_id": "$AGENT_ID",
+  "start_time": "$(date -Iseconds)",
+  "process_id": "$$",
+  "memory_limit": "$MEMORY_LIMIT",
+  "cpu_limit": "$CPU_LIMIT",
+  "timeout": "$TIMEOUT",
+  "samples": []
+}
+EOF
+}
+
+# Collect process metrics
+collect_metrics() {
+    local pid="${1:-$$}"
+    local timestamp="$(date -Iseconds)"
+
+    # Get process statistics
+    local mem_usage=""
+    local cpu_usage=""
+    local open_files=""
+    local threads=""
+
+    if command -v ps >/dev/null 2>&1; then
+        mem_usage=$(ps -p "$pid" -o rss= 2>/dev/null | tr -d ' ' || echo "0")
+        cpu_usage=$(ps -p "$pid" -o %cpu= 2>/dev/null | tr -d ' ' || echo "0")
+    fi
+
+    if command -v lsof >/dev/null 2>&1; then
+        open_files=$(lsof -p "$pid" 2>/dev/null | wc -l || echo "0")
+    fi
+
+    if [[ -f "/proc/$pid/status" ]]; then
+        threads=$(grep "^Threads:" "/proc/$pid/status" | awk '{print $2}' || echo "0")
+    fi
+
+    # Create metrics entry
+    local metrics_entry=$(cat << EOF
+{
+  "timestamp": "$timestamp",
+  "memory_kb": "$mem_usage",
+  "cpu_percent": "$cpu_usage",
+  "open_files": "$open_files",
+  "threads": "$threads"
+}
+EOF
+    )
+
+    # Update metrics file
+    if [[ -f "$METRICS_FILE" ]]; then
+        # Use jq to safely append to samples array
+        if command -v jq >/dev/null 2>&1; then
+            jq --argjson entry "$metrics_entry" '.samples += [$entry]' "$METRICS_FILE" > "$METRICS_FILE.tmp" && \
+                mv "$METRICS_FILE.tmp" "$METRICS_FILE"
+        else
+            # Fallback without jq
+            echo "Warning: jq not available, using simple append" >&2
+            echo "$metrics_entry" >> "$METRICS_FILE.raw"
+        fi
+    fi
+}
+
+# Check resource limits
+check_limits() {
+    local pid="${1:-$$}"
+
+    # Memory limit check
+    if command -v ps >/dev/null 2>&1; then
+        local mem_kb=$(ps -p "$pid" -o rss= 2>/dev/null | tr -d ' ' || echo "0")
+        local mem_mb=$((mem_kb / 1024))
+
+        case "$MEMORY_LIMIT" in
+            *G|*g)
+                local limit_mb=$((${MEMORY_LIMIT%[Gg]*} * 1024))
+                ;;
+            *M|*m)
+                local limit_mb=$((${MEMORY_LIMIT%[Mm]*}))
+                ;;
+            *)
+                local limit_mb=2048  # Default 2GB
+                ;;
+        esac
+
+        if [[ $mem_mb -gt $limit_mb ]]; then
+            log_warning "Memory limit exceeded: ${mem_mb}MB > ${limit_mb}MB"
+            return 1
+        fi
+    fi
+
+    # CPU limit check
+    if command -v ps >/dev/null 2>&1; then
+        local cpu_percent=$(ps -p "$pid" -o %cpu= 2>/dev/null | tr -d ' ' || echo "0")
+        local cpu_limit_num=$((${CPU_LIMIT%\%}))
+
+        if (( $(echo "$cpu_percent > $cpu_limit_num" | bc -l) )); then
+            log_warning "CPU limit exceeded: ${cpu_percent}% > ${CPU_LIMIT}"
+            return 1
+        fi
+    fi
+
+    return 0
+}
+
+# Start background monitoring
+start_monitoring() {
+    local pid="${1:-$$}"
+    local interval="${2:-30}"  # Check every 30 seconds
+
+    log_info "Starting process monitoring for PID $pid (interval: ${interval}s)"
+
+    # Start monitoring in background
+    (
+        while true; do
+            if ! kill -0 "$pid" 2>/dev/null; then
+                log_info "Process $pid no longer exists, stopping monitoring"
+                break
+            fi
+
+            collect_metrics "$pid"
+
+            if ! check_limits "$pid"; then
+                log_error "Resource limits exceeded, terminating process $pid"
+                kill -TERM "$pid" 2>/dev/null || true
+                break
+            fi
+
+            sleep "$interval"
+        done
+    ) &
+
+    MONITOR_PID=$!
+    echo "$MONITOR_PID"
+}
+
+# Stop monitoring
+stop_monitoring() {
+    if [[ -n "$MONITOR_PID" ]] && kill -0 "$MONITOR_PID" 2>/dev/null; then
+        kill "$MONITOR_PID" 2>/dev/null || true
+        log_info "Stopped monitoring (PID: $MONITOR_PID)"
+    fi
+}
+
+# Generate final report
+generate_report() {
+    local exit_code="${1:-0}"
+
+    if [[ -f "$METRICS_FILE" ]]; then
+        # Update with final information
+        if command -v jq >/dev/null 2>&1; then
+            jq --arg end_time "$(date -Iseconds)" \
+               --arg exit_code "$exit_code" \
+               '.end_time = $end_time | .exit_code = $exit_code' \
+               "$METRICS_FILE" > "$METRICS_FILE.tmp" && \
+               mv "$METRICS_FILE.tmp" "$METRICS_FILE"
+        fi
+
+        log_success "Process report generated: $METRICS_FILE"
+
+        # Print summary
+        if command -v jq >/dev/null 2>&1; then
+            local samples=$(jq '.samples | length' "$METRICS_FILE")
+            echo "📊 Process Metrics Summary:" >&2
+            echo "  Agent ID: $AGENT_ID" >&2
+            echo "  Samples: $samples" >&2
+            echo "  Exit Code: $exit_code" >&2
+        fi
+    fi
+}
+
+# Monitor existing process
+monitor_pid() {
+    local pid="$1"
+
+    log_info "Monitoring existing process: PID $pid"
+
+    if ! kill -0 "$pid" 2>/dev/null; then
+        log_error "Process $pid does not exist"
+        return 1
+    fi
+
+    # Start monitoring
+    local monitor_pid=$(start_monitoring "$pid")
+
+    # Wait for process to complete
+    while kill -0 "$pid" 2>/dev/null; do
+        sleep 5
+    done
+
+    # Stop monitoring
+    stop_monitoring
+
+    log_success "Process monitoring completed for PID $pid"
+}
+
+# Main execution
+main() {
+    local action="${1:-"instrument"}"
+
+    case "$action" in
+        "instrument")
+            init_telemetry
+            local monitor_pid=$(start_monitoring)
+
+            # Set up cleanup traps
+            trap 'stop_monitoring; generate_report $?' EXIT
+            trap 'stop_monitoring; generate_report 1' INT TERM
+
+            log_success "Process instrumentation started for $AGENT_ID"
+            ;;
+        "monitor-pid")
+            if [[ -z "${2:-}" ]]; then
+                log_error "PID required for monitor-pid action"
+                exit 1
+            fi
+            monitor_pid "$2"
+            ;;
+        "--help"|"-h")
+            cat << EOF
+CFN Process Instrumentation Script
+
+Usage:
+  $0                     # Instrument current process
+  $0 monitor-pid <pid>   # Monitor existing process
+  $0 --help             # Show this help
+
+Environment Variables:
+  AGENT_ID              # Agent identifier (default: hostname-PID)
+  CFN_MEMORY_LIMIT      # Memory limit (default: 2G)
+  CFN_CPU_LIMIT         # CPU limit (default: 80%)
+  CFN_TIMEOUT           # Timeout in seconds (default: 600)
+  CFN_TELEMETRY_DIR     # Telemetry storage directory
+
+This script provides process monitoring and resource limit enforcement
+for CFN Loop agents and orchestration processes.
+EOF
+            ;;
+        *)
+            log_error "Unknown action: $action"
+            exit 1
+            ;;
+    esac
+}
+
+# Execute main function if run directly
+if [[ "${BASH_SOURCE[0]}" == "${0}" ]]; then
+    main "$@"
+else
+    # When sourced, automatically instrument current process
+    init_telemetry
+    local monitor_pid=$(start_monitoring)
+
+    # Set up cleanup traps
+    trap 'stop_monitoring; generate_report $?' EXIT
+    trap 'stop_monitoring; generate_report 1' INT TERM
+
+    log_info "Process instrumentation enabled for $AGENT_ID"
+fi

package/claude-assets/skills/cfn-product-owner-decision/execute-decision.sh

@@ -77,10 +77,49 @@ echo -e "${YELLOW}📥 Retrieving Loop 2 context...${NC}"
 LOOP2_FEEDBACK=$(redis-cli HGET "swarm:${TASK_ID}:loop2:consensus" "feedback" || echo "")
 TASK_CONTEXT=$(redis-cli HGETALL "swarm:${TASK_ID}:context" || echo "")
 
-# Build Product Owner context
+# Retrieve comprehensive audit trail data
+echo -e "${YELLOW}🔍 Retrieving audit trail data for informed decision-making...${NC}"
+set +e
+AUDIT_DATA=$(./.claude/skills/cfn-task-audit/get-audit-data.sh \
+  --task-id "$TASK_ID" \
+  --mode combined \
+  --format json 2>/dev/null || echo "[]")
+
+AUDIT_SUMMARY=$(./.claude/skills/cfn-task-audit/get-audit-data.sh \
+  --task-id "$TASK_ID" \
+  --mode combined \
+  --format summary 2>/dev/null || echo "No audit data available")
+set -e
+
+# Analyze audit data for patterns
+AUDIT_INSIGHTS=""
+if [ "$AUDIT_DATA" != "[]" ]; then
+  # Extract key patterns from audit data
+  PREVIOUS_DECISIONS=$(echo "$AUDIT_DATA" | jq -r '.[] | select(.agent_type == "product-owner") | .decision' 2>/dev/null | tr '\n' ', ' | sed 's/,$//' || echo "None")
+  AGENT_PERFORMANCE=$(echo "$AUDIT_DATA" | jq -r 'group_by(.agent_type) | map({agent: .[0].agent_type, avg_confidence: map(.confidence) | add / length}) | sort_by(.avg_confidence) | reverse | .[0:3] | .[] | "\(.agent): \(.avg_confidence)"' 2>/dev/null || echo "No performance data")
+
+  # Check for repeating concerns
+  REPEATING_CONCERNS=$(echo "$AUDIT_DATA" | jq -r '.[] | select(.agent_type == "reviewer" or .agent_type == "tester") | .reasoning | scan("security|performance|scope|quality|bug")' 2>/dev/null | sort | uniq -c | sort -nr | head -3 | awk '{print $2 " (" $1 "x)"}' | tr '\n' ', ' | sed 's/,$//' || echo "No repeating concerns")
+
+  AUDIT_INSIGHTS="
+AUDIT TRAIL INSIGHTS:
+- Previous Product Owner Decisions: $PREVIOUS_DECISIONS
+- Top Performing Agents: $AGENT_PERFORMANCE
+- Repeating Concerns: $REPEATING_CONCERNS
+- Total Audit Records: $(echo "$AUDIT_DATA" | jq '. | length' 2>/dev/null || echo "0")
+
+Audit Summary:
+$AUDIT_SUMMARY
+"
+else
+  AUDIT_INSIGHTS="AUDIT TRAIL: No historical data available for this task."
+fi
+
+# Build enhanced Product Owner context with audit insights
 PO_CONTEXT="
 You are the Product Owner making a strategic decision for CFN Loop iteration $ITERATION of $MAX_ITERATIONS.
 
+CURRENT ITERATION DATA:
 Loop 2 Consensus: $CONSENSUS
 Threshold: $THRESHOLD
 Success Criteria: ${SUCCESS_CRITERIA:-"Not specified"}
@@ -91,15 +130,26 @@ $LOOP2_FEEDBACK
 Task Context:
 $TASK_CONTEXT
 
-Make a strategic decision:
-- PROCEED: Quality threshold met, deliverables complete
-- ITERATE: Improvements needed, iterations remaining
-- ABORT: Max iterations reached or unrecoverable failure
+$AUDIT_INSIGHTS
+
+ENHANCED DECISION FRAMEWORK:
+Use the audit trail insights to inform your decision. Consider:
+- Are there repeating concerns that suggest systematic issues?
+- Which agents have performed well on similar tasks?
+- Do previous decisions show a pattern of success or failure?
+- Are there cross-mode inconsistencies that need attention?
+
+DECISION OPTIONS:
+- PROCEED: Quality threshold met, deliverables complete, audit shows positive trajectory
+- ITERATE: Improvements needed, iterations remaining, audit shows recoverable issues
+- ABORT: Max iterations reached, systematic failure, or audit shows insurmountable barriers
 
 Output format:
 Decision: PROCEED|ITERATE|ABORT
-Reasoning: [your explanation]
+Reasoning: [your explanation + audit insights]
 Confidence: [0.0-1.0]
+Audit Analysis: [brief summary of how audit data influenced your decision]
+Agent Performance: [any observations about agent reliability from audit trail]
 "
 
 # Spawn Product Owner agent
@@ -144,6 +194,12 @@ else
 
     # Parse confidence
     CONFIDENCE=$(echo "$PO_OUTPUT" | grep -oE "Confidence:\s*[0-9]+\.?[0-9]*" | grep -oE "[0-9]+\.?[0-9]*" || echo "0.85")
+
+    # Parse audit analysis (enhanced output)
+    AUDIT_ANALYSIS=$(echo "$PO_OUTPUT" | grep -oiE "Audit Analysis:\s*.*" | sed 's/Audit Analysis:\s*//' || echo "No audit analysis provided")
+
+    # Parse agent performance observations
+    AGENT_PERFORMANCE_OBSERVATIONS=$(echo "$PO_OUTPUT" | grep -oiE "Agent Performance:\s*.*" | sed 's/Agent Performance:\s*//' || echo "No agent performance observations")
   else
     echo -e "${RED}❌ ERROR: Product Owner output file missing or empty${NC}"
     echo "Expected: $PO_OUTPUT_FILE"
@@ -167,6 +223,11 @@ fi
 echo -e "${GREEN}✅ Product Owner Decision: $DECISION_TYPE${NC}"
 echo "Reasoning: $REASONING"
 echo "Confidence: $CONFIDENCE"
+if [ "$AUDIT_DATA" != "[]" ]; then
+  echo -e "${BLUE}📊 Audit Analysis: $AUDIT_ANALYSIS${NC}"
+  echo -e "${BLUE}🏆 Agent Performance: $AGENT_PERFORMANCE_OBSERVATIONS${NC}"
+  echo -e "${BLUE}📈 Audit Records Analyzed: $(echo "$AUDIT_DATA" | jq '. | length' 2>/dev/null || echo "0")${NC}"
+fi
 
 # Deliverable verification for PROCEED decisions
 if [ "$DECISION_TYPE" = "PROCEED" ]; then
@@ -254,7 +315,7 @@ else
   echo -e "${GREEN}No deferred items detected in Product Owner output${NC}"
 fi
 
-# Build decision JSON
+# Build enhanced decision JSON with audit insights
 DECISION_JSON=$(cat <<EOF
 {
   "decision": "$DECISION_TYPE",
@@ -263,17 +324,25 @@ DECISION_JSON=$(cat <<EOF
   "iteration": $ITERATION,
   "consensus": $CONSENSUS,
   "threshold": $THRESHOLD,
-  "timestamp": $(date +%s)
+  "timestamp": $(date +%s),
+  "audit_analysis": "$AUDIT_ANALYSIS",
+  "agent_performance_observations": "$AGENT_PERFORMANCE_OBSERVATIONS",
+  "audit_records_analyzed": $(echo "$AUDIT_DATA" | jq '. | length' 2>/dev/null || echo "0"),
+  "audit_informed": $(if [ "$AUDIT_DATA" != "[]" ]; then echo "true"; else echo "false"; fi)
 }
 EOF
 )
 
-# Store decision in Redis
-echo -e "${YELLOW}💾 Storing decision in Redis...${NC}"
+# Store decision in Redis with audit context
+echo -e "${YELLOW}💾 Storing decision in Redis with audit context...${NC}"
 redis-cli LPUSH "swarm:${TASK_ID}:decision" "$DECISION_TYPE"
 redis-cli HSET "swarm:${TASK_ID}:${AGENT_ID}:result" "decision" "$DECISION_TYPE"
 redis-cli HSET "swarm:${TASK_ID}:${AGENT_ID}:result" "reasoning" "$REASONING"
 redis-cli HSET "swarm:${TASK_ID}:${AGENT_ID}:result" "confidence" "$CONFIDENCE"
+redis-cli HSET "swarm:${TASK_ID}:${AGENT_ID}:result" "audit_analysis" "$AUDIT_ANALYSIS"
+redis-cli HSET "swarm:${TASK_ID}:${AGENT_ID}:result" "agent_performance_observations" "$AGENT_PERFORMANCE_OBSERVATIONS"
+redis-cli HSET "swarm:${TASK_ID}:${AGENT_ID}:result" "audit_records_analyzed" "$(echo "$AUDIT_DATA" | jq '. | length' 2>/dev/null || echo "0")"
+redis-cli HSET "swarm:${TASK_ID}:${AGENT_ID}:result" "audit_informed" "$(if [ "$AUDIT_DATA" != "[]" ]; then echo "true"; else echo "false"; fi)"
 
 # Store in metrics
 redis-cli LPUSH "swarm:${TASK_ID}:metrics:product_owner_decisions" "$DECISION_JSON"
@@ -293,3 +362,6 @@ redis-cli LPUSH "swarm:${TASK_ID}:${AGENT_ID}:done" "complete"
 echo "$DECISION_JSON"
 
 echo -e "${GREEN}✅ Product Owner decision execution complete${NC}"
+if [ "$AUDIT_DATA" != "[]" ]; then
+  echo -e "${BLUE}📊 Decision was informed by audit trail analysis${NC}"
+fi