claude-flow-novice 2.15.3 → 2.15.4

package/claude-assets/skills/cfn-loop-orchestration/helpers/gate-check.sh

@@ -1,27 +1,51 @@
 #!/usr/bin/env bash
 
 ##############################################################################
-# Gate Check Helper
-# Validates Loop 3 self-assessment against gate threshold
+# Gate Check Helper - Test-Driven Validation with Hybrid Fallback
+# Validates Loop 3 self-assessment using test-driven or confidence-based strategies
 #
 # Usage:
 #   gate-check.sh --task-id <id> \
 #                 --agents <agent1,agent2,...> \
 #                 --threshold <0.0-1.0> \
-#                 --min-quorum <n|n%|0.n>
+#                 --min-quorum <n|n%|0.n> \
+#                 [--mode <mvp|standard|enterprise>] \
+#                 [--success-criteria <json>] \
+#                 [--strategy <test-driven|confidence|auto>]
+#
+# Environment:
+#   CFN_GATE_STRATEGY: test-driven|confidence|auto (default: auto)
 #
 # Returns:
-#   Exit 0: Gate passed (Loop 3 meets threshold)
-#   Exit 1: Gate failed (Loop 3 needs iteration)
+#   Exit 0: Gate passed
+#   Exit 1: Gate failed (needs iteration)
 ##############################################################################
 
 set -euo pipefail
 
+# Script directory for helper resolution
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PROJECT_ROOT="$(cd "$SCRIPT_DIR/../../.." && pwd)"
+
+# SECURITY FIX #1: Path Traversal Prevention (CWE-22)
+# Validate PROJECT_ROOT is within expected location to prevent symlink attacks
+EXPECTED_PREFIX="/home/user/claude-flow-novice"
+if [[ ! "$PROJECT_ROOT" =~ ^${EXPECTED_PREFIX//./\\.} ]]; then
+    echo "❌ SECURITY ERROR: Invalid project root detected" >&2
+    echo "   Expected prefix: $EXPECTED_PREFIX" >&2
+    echo "   Actual path: $PROJECT_ROOT" >&2
+    echo "   Risk: Path traversal / symlink attack" >&2
+    exit 1
+fi
+
 # Parameters
 TASK_ID=""
 AGENTS=""
 THRESHOLD=""
 MIN_QUORUM=""
+MODE="standard"
+SUCCESS_CRITERIA=""
+STRATEGY="${CFN_GATE_STRATEGY:-auto}"
 
 # Parse arguments
 while [[ $# -gt 0 ]]; do
@@ -30,61 +54,541 @@ while [[ $# -gt 0 ]]; do
     --agents) AGENTS="$2"; shift 2 ;;
     --threshold) THRESHOLD="$2"; shift 2 ;;
     --min-quorum) MIN_QUORUM="$2"; shift 2 ;;
+    --mode) MODE="$2"; shift 2 ;;
+    --success-criteria) SUCCESS_CRITERIA="$2"; shift 2 ;;
+    --strategy) STRATEGY="$2"; shift 2 ;;
     *) echo "Unknown option: $1"; exit 1 ;;
   esac
 done
 
 # Validation
 if [ -z "$TASK_ID" ] || [ -z "$AGENTS" ] || [ -z "$THRESHOLD" ] || [ -z "$MIN_QUORUM" ]; then
-  echo "Error: Missing required parameters"
+  echo "Error: Missing required parameters" >&2
   exit 1
 fi
 
-# Use Redis Coordination skill to collect confidence scores
-SKILL_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.claude/skills/cfn-cfn-.." && pwd)"
-REDIS_COORD_SKILL="$SKILL_DIR/redis-coordination"
+##############################################################################
+# Helper Functions
+##############################################################################
 
-echo "Gate Check Configuration:"
-echo "  Task ID: $TASK_ID"
-echo "  Agent IDs: $AGENTS"
-echo "  Min Quorum: $MIN_QUORUM"
-echo ""
+# Validate JSON structure of success criteria
+# SECURITY FIX #2: JSON Schema Validation (CWE-400)
+# Prevents DoS via excessively large test suite arrays and invalid field values
+validate_success_criteria() {
+  local CRITERIA="$1"
+  local MAX_TEST_SUITES=50  # DoS prevention: max 50 test suites
+  local MAX_FIELD_LENGTH=256  # Field name length limit
+  local PASS_THRESHOLD_MIN=0.0
+  local PASS_THRESHOLD_MAX=1.0
+  local TIMEOUT_MIN=1
+  local TIMEOUT_MAX=3600
 
-# Collect Loop 3 confidence scores
-# Note: invoke-waiting-mode.sh outputs consensus to stdout and verbose messages to stderr
-# We capture only stdout to get the numeric consensus value
-CONSENSUS=$("$REDIS_COORD_SKILL/invoke-waiting-mode.sh" collect \
-  --task-id "$TASK_ID" \
-  --agent-ids "$AGENTS" \
-  --min-quorum "$MIN_QUORUM") || {
-  echo "❌ Error: Failed to collect Loop 3 confidence scores"
-  echo "   Agent IDs: $AGENTS"
-  echo "   Output: $CONSENSUS"
-  exit 1
+  if [ -z "$CRITERIA" ]; then
+    echo "❌ No success criteria provided" >&2
+    return 1
+  fi
+
+  # Validate JSON structure
+  if ! echo "$CRITERIA" | jq empty 2>/dev/null; then
+    echo "❌ Invalid JSON in success criteria" >&2
+    return 1
+  fi
+
+  # Check for required fields
+  if ! echo "$CRITERIA" | jq -e '.test_suites' >/dev/null 2>&1; then
+    echo "❌ Missing test_suites array in success criteria" >&2
+    return 1
+  fi
+
+  # SECURITY FIX #2a: Array size validation - prevent DoS
+  local TEST_SUITE_COUNT
+  TEST_SUITE_COUNT=$(echo "$CRITERIA" | jq '.test_suites | length')
+
+  if [ -z "$TEST_SUITE_COUNT" ] || [ "$TEST_SUITE_COUNT" -lt 0 ]; then
+    echo "❌ Invalid test_suites array" >&2
+    return 1
+  fi
+
+  if [ "$TEST_SUITE_COUNT" -gt "$MAX_TEST_SUITES" ]; then
+    echo "❌ SECURITY ERROR: test_suites array exceeds maximum size" >&2
+    echo "   Count: $TEST_SUITE_COUNT (max: $MAX_TEST_SUITES)" >&2
+    echo "   Risk: DoS via resource exhaustion" >&2
+    return 1
+  fi
+
+  # SECURITY FIX #2b: Field-level validation
+  local SUITE_INDEX=0
+  while [ "$SUITE_INDEX" -lt "$TEST_SUITE_COUNT" ]; do
+    local SUITE
+    SUITE=$(echo "$CRITERIA" | jq ".test_suites[$SUITE_INDEX]")
+
+    # Validate pass_threshold (0.0-1.0)
+    local PASS_THRESHOLD
+    PASS_THRESHOLD=$(echo "$SUITE" | jq -r '.pass_threshold // 0.5')
+
+    if ! [[ "$PASS_THRESHOLD" =~ ^[0-9]+\.?[0-9]*$ ]]; then
+      echo "❌ Invalid pass_threshold in test suite $SUITE_INDEX: $PASS_THRESHOLD" >&2
+      return 1
+    fi
+
+    # Use bc for floating point comparison
+    if (( $(echo "$PASS_THRESHOLD < $PASS_THRESHOLD_MIN" | bc -l) )) || \
+       (( $(echo "$PASS_THRESHOLD > $PASS_THRESHOLD_MAX" | bc -l) )); then
+      echo "❌ pass_threshold out of range [0.0-1.0] in suite $SUITE_INDEX: $PASS_THRESHOLD" >&2
+      return 1
+    fi
+
+    # SECURITY FIX #2c: Timeout range validation (1-3600 seconds)
+    local TIMEOUT
+    TIMEOUT=$(echo "$SUITE" | jq -r '.timeout // 300')
+
+    if ! [[ "$TIMEOUT" =~ ^[0-9]+$ ]]; then
+      echo "❌ Invalid timeout in test suite $SUITE_INDEX: $TIMEOUT (must be integer)" >&2
+      return 1
+    fi
+
+    if [ "$TIMEOUT" -lt "$TIMEOUT_MIN" ] || [ "$TIMEOUT" -gt "$TIMEOUT_MAX" ]; then
+      echo "❌ Timeout out of range [${TIMEOUT_MIN}-${TIMEOUT_MAX}s] in suite $SUITE_INDEX: ${TIMEOUT}s" >&2
+      return 1
+    fi
+
+    # Validate suite name length
+    local SUITE_NAME
+    SUITE_NAME=$(echo "$SUITE" | jq -r '.name // "unnamed"')
+
+    if [ ${#SUITE_NAME} -gt "$MAX_FIELD_LENGTH" ]; then
+      echo "❌ Test suite name exceeds maximum length in suite $SUITE_INDEX" >&2
+      return 1
+    fi
+
+    SUITE_INDEX=$((SUITE_INDEX + 1))
+  done
+
+  return 0
 }
 
-# Validate consensus is a valid number
-if ! [[ "$CONSENSUS" =~ ^[0-9]+\.?[0-9]*$ ]]; then
-  echo "⚠️  WARNING: Invalid consensus value: $CONSENSUS (expected numeric)"
-  echo "   Defaulting to 0.0"
-  CONSENSUS="0.0"
-fi
+# Check if success criteria exists and is valid
+has_success_criteria() {
+  if [ -z "$SUCCESS_CRITERIA" ]; then
+    return 1
+  fi
 
-echo "Loop 3 Gate Check:"
-echo "  Consensus: $CONSENSUS"
-echo "  Threshold: $THRESHOLD"
-echo "  Required: >= $THRESHOLD"
+  validate_success_criteria "$SUCCESS_CRITERIA" >/dev/null 2>&1
+}
 
-# Compare consensus to gate threshold
-if (( $(echo "$CONSENSUS >= $THRESHOLD" | bc -l) )); then
-  echo "✅ Gate PASSED - Loop 3 self-validation successful"
+# Get mode-specific test pass threshold
+get_mode_threshold() {
+  local MODE="$1"
 
-  # NOTE: Orchestrator controls Loop 2 spawn timing directly (orchestrate.sh line 520)
-  # No signal broadcast needed - Loop 2 agents spawn after this helper returns 0
+  case "$MODE" in
+    mvp) echo "0.80" ;;
+    standard) echo "0.95" ;;
+    enterprise) echo "0.99" ;;
+    *) echo "0.95" ;;  # Default to standard
+  esac
+}
 
-  exit 0
-else
-  echo "❌ Gate FAILED - Loop 3 needs improvement"
-  echo "   Gap: $(echo "$THRESHOLD - $CONSENSUS" | bc -l)"
-  exit 1
-fi
+# Validate command safety (prevent shell injection)
+validate_command_safety() {
+  local COMMAND="$1"
+
+  # Check for dangerous shell patterns
+  # Allow: && (AND operator), || (OR operator)
+  # Block: ; (command separator), | (pipe), > < (redirects), ` (backticks), $() (command substitution), {} (brace expansion)
+
+  # Remove safe operators first
+  local SANITIZED="${COMMAND//&&/}"
+  SANITIZED="${SANITIZED//||/}"
+
+  # Now check for remaining dangerous metacharacters
+  if [[ "$SANITIZED" =~ [\;\|\>\<\`\$\(\)\{\}] ]]; then
+    echo "❌ Unsafe command detected: contains dangerous shell metacharacters" >&2
+    echo "   Blocked patterns: ; | > < \` \$() {}" >&2
+    return 1
+  fi
+
+  return 0
+}
+
+# Check if command is required or optional
+is_required() {
+  local SUITE_JSON="$1"
+  local REQUIRED=$(echo "$SUITE_JSON" | jq -r '.required // true')
+
+  [ "$REQUIRED" = "true" ]
+}
+
+# Execute a single test suite
+execute_test_suite() {
+  local SUITE_JSON="$1"
+  local SUITE_NAME=$(echo "$SUITE_JSON" | jq -r '.name // "unnamed"')
+  local COMMAND=$(echo "$SUITE_JSON" | jq -r '.command')
+  local TIMEOUT=$(echo "$SUITE_JSON" | jq -r '.timeout // 300')
+  local FRAMEWORK=$(echo "$SUITE_JSON" | jq -r '.framework // "auto"')
+
+  echo "  Executing test suite: $SUITE_NAME" >&2
+
+  # Validate command safety
+  if ! validate_command_safety "$COMMAND"; then
+    echo "    ❌ Command validation failed" >&2
+    return 1
+  fi
+
+  # Execute with timeout
+  local OUTPUT
+  local EXIT_CODE=0
+
+  OUTPUT=$(cd "$PROJECT_ROOT" && timeout "$TIMEOUT" bash -c "$COMMAND" 2>&1) || EXIT_CODE=$?
+
+  if [ $EXIT_CODE -eq 124 ]; then
+    echo "    ⚠️  Test execution timed out after ${TIMEOUT}s" >&2
+    echo '{"pass_rate": 0.0, "passed": 0, "failed": 0, "total": 0, "status": "timeout"}'
+    return 1
+  elif [ $EXIT_CODE -ne 0 ]; then
+    echo "    ⚠️  Test execution failed with exit code $EXIT_CODE" >&2
+  fi
+
+  # Parse test results
+  local RESULTS
+  if [ -x "$PROJECT_ROOT/.claude/skills/cfn-loop-orchestration/helpers/parse-test-results.sh" ]; then
+    # Use dedicated parser if available (expects positional params: framework, output)
+    RESULTS=$("$PROJECT_ROOT/.claude/skills/cfn-loop-orchestration/helpers/parse-test-results.sh" \
+      "$FRAMEWORK" \
+      "$OUTPUT" 2>/dev/null) || {
+      echo "    ⚠️  Failed to parse test results" >&2
+      RESULTS='{"pass_rate": 0.0, "passed": 0, "failed": 0, "total": 0, "status": "parse_error"}'
+    }
+  else
+    # Fallback: simple pattern matching
+    RESULTS=$(parse_test_results_fallback "$OUTPUT" "$FRAMEWORK" "$EXIT_CODE")
+  fi
+
+  local PASS_RATE=$(echo "$RESULTS" | jq -r '.pass_rate // 0.0')
+  local PASSED=$(echo "$RESULTS" | jq -r '.passed // 0')
+  local TOTAL=$(echo "$RESULTS" | jq -r '.total // 0')
+
+  echo "    Pass rate: $PASS_RATE ($PASSED/$TOTAL tests)" >&2
+
+  echo "$RESULTS"
+}
+
+# Fallback test result parser (simple pattern matching)
+parse_test_results_fallback() {
+  local OUTPUT="$1"
+  local FRAMEWORK="$2"
+  local EXIT_CODE="$3"
+
+  local PASSED=0
+  local FAILED=0
+  local TOTAL=0
+
+  # Try common test output patterns
+  if echo "$OUTPUT" | grep -qE "([0-9]+) passed.*([0-9]+) failed"; then
+    PASSED=$(echo "$OUTPUT" | grep -oE "([0-9]+) passed" | grep -oE "[0-9]+" | head -1)
+    FAILED=$(echo "$OUTPUT" | grep -oE "([0-9]+) failed" | grep -oE "[0-9]+" | head -1)
+  elif echo "$OUTPUT" | grep -qE "([0-9]+)/([0-9]+) tests? passed"; then
+    PASSED=$(echo "$OUTPUT" | grep -oE "([0-9]+)/([0-9]+)" | cut -d'/' -f1)
+    TOTAL=$(echo "$OUTPUT" | grep -oE "([0-9]+)/([0-9]+)" | cut -d'/' -f2)
+    FAILED=$((TOTAL - PASSED))
+  fi
+
+  TOTAL=$((PASSED + FAILED))
+
+  # Calculate pass rate
+  local PASS_RATE="0.0"
+  if [ $TOTAL -gt 0 ]; then
+    PASS_RATE=$(echo "scale=2; $PASSED / $TOTAL" | bc -l)
+  elif [ $EXIT_CODE -eq 0 ]; then
+    # No test output but exit code 0 - assume success
+    PASS_RATE="1.0"
+    PASSED=1
+    TOTAL=1
+  fi
+
+  echo "{\"pass_rate\": $PASS_RATE, \"passed\": $PASSED, \"failed\": $FAILED, \"total\": $TOTAL, \"status\": \"parsed\"}"
+}
+
+# Calculate aggregate pass rate from multiple test suites
+calculate_aggregate_pass_rate() {
+  local RESULTS_FILE="$1"
+
+  local TOTAL_PASSED=0
+  local TOTAL_TESTS=0
+
+  while IFS= read -r RESULT; do
+    local PASSED=$(echo "$RESULT" | jq -r '.passed // 0')
+    local TOTAL=$(echo "$RESULT" | jq -r '.total // 0')
+
+    TOTAL_PASSED=$((TOTAL_PASSED + PASSED))
+    TOTAL_TESTS=$((TOTAL_TESTS + TOTAL))
+  done < "$RESULTS_FILE"
+
+  if [ $TOTAL_TESTS -eq 0 ]; then
+    echo "0.0"
+  else
+    echo "scale=4; $TOTAL_PASSED / $TOTAL_TESTS" | bc -l
+  fi
+}
+
+# Store test results (Redis for CLI mode, file for Task mode)
+store_test_results() {
+  local TASK_ID="$1"
+  local PASS_RATE="$2"
+  local RESULTS_FILE="$3"
+
+  # Check if running in CLI mode (Redis available)
+  if command -v redis-cli >/dev/null 2>&1 && redis-cli ping >/dev/null 2>&1; then
+    # Store in Redis
+    redis-cli HSET "task:$TASK_ID:gate" "test_pass_rate" "$PASS_RATE" >/dev/null
+    redis-cli HSET "task:$TASK_ID:gate" "test_results" "$(cat "$RESULTS_FILE")" >/dev/null
+    redis-cli EXPIRE "task:$TASK_ID:gate" 86400 >/dev/null  # 24h TTL
+  else
+    # Task mode: store in temp file for coordinator to read
+    local OUTPUT_DIR="/tmp/cfn-gate-results"
+    mkdir -p "$OUTPUT_DIR"
+    echo "$PASS_RATE" > "$OUTPUT_DIR/$TASK_ID.pass_rate"
+    cp "$RESULTS_FILE" "$OUTPUT_DIR/$TASK_ID.results.json"
+  fi
+}
+
+# Generate iteration context for failed gate
+generate_iteration_context() {
+  local TASK_ID="$1"
+  local PASS_RATE="$2"
+  local THRESHOLD="$3"
+  local RESULTS_FILE="$4"
+
+  local CONTEXT_FILE="/tmp/cfn-iteration-context-$TASK_ID.json"
+
+  # Extract failed test details
+  local FAILED_TESTS=$(jq -s '[.[] | select(.pass_rate < 1.0)]' "$RESULTS_FILE")
+
+  # Generate context
+  cat > "$CONTEXT_FILE" <<EOF
+{
+  "gate_status": "failed",
+  "pass_rate": $PASS_RATE,
+  "threshold": $THRESHOLD,
+  "gap": $(echo "$THRESHOLD - $PASS_RATE" | bc -l),
+  "failed_tests": $FAILED_TESTS,
+  "recommendations": [
+    "Review failed test suites",
+    "Fix implementation issues",
+    "Re-run validation"
+  ]
+}
+EOF
+
+  echo "  Iteration context saved to: $CONTEXT_FILE" >&2
+}
+
+##############################################################################
+# Test-Driven Gate Check
+##############################################################################
+
+gate_check_test_driven() {
+  local TASK_ID="$1"
+  local MODE="$2"
+  local SUCCESS_CRITERIA="$3"
+
+  # SECURITY FIX #3: DoS Prevention - Total Time Limit
+  # Prevent unbounded test execution time (default 30 min, configurable)
+  local MAX_TOTAL_TIME=${CFN_MAX_GATE_TIME:-1800}  # Default 30 minutes (1800 seconds)
+  local START_TIME=$(date +%s)
+
+  echo "🧪 Test-Driven Gate Check"
+  echo "  Task ID: $TASK_ID"
+  echo "  Mode: $MODE"
+  echo "  Max Total Time: ${MAX_TOTAL_TIME}s"
+  echo ""
+
+  # Validate success criteria
+  if ! validate_success_criteria "$SUCCESS_CRITERIA"; then
+    echo "❌ Invalid success criteria" >&2
+    return 1
+  fi
+
+  # Extract test suites
+  local TEST_SUITES=$(echo "$SUCCESS_CRITERIA" | jq -c '.test_suites[]')
+
+  if [ -z "$TEST_SUITES" ]; then
+    echo "❌ No test suites defined in success criteria" >&2
+    return 1
+  fi
+
+  # Temporary file for results
+  # SECURITY FIX #4: Secure Temp File Permissions
+  # Set restrictive permissions (owner read/write only) to prevent information disclosure
+  local RESULTS_FILE=$(mktemp)
+  chmod 600 "$RESULTS_FILE"
+  trap "rm -f '$RESULTS_FILE'" EXIT
+
+  # Execute each test suite
+  local SUITE_COUNT=0
+  local FAILED_REQUIRED=0
+
+  while IFS= read -r SUITE; do
+    SUITE_COUNT=$((SUITE_COUNT + 1))
+
+    # SECURITY FIX #3 (continued): Check total execution time
+    local CURRENT_TIME=$(date +%s)
+    local ELAPSED=$((CURRENT_TIME - START_TIME))
+
+    if [ $ELAPSED -gt $MAX_TOTAL_TIME ]; then
+      echo "❌ SECURITY ERROR: Total execution time exceeded" >&2
+      echo "   Elapsed: ${ELAPSED}s > Max: ${MAX_TOTAL_TIME}s" >&2
+      echo "   Risk: DoS via unbounded execution" >&2
+      return 1
+    fi
+
+    local RESULT
+    RESULT=$(execute_test_suite "$SUITE") || {
+      if is_required "$SUITE"; then
+        FAILED_REQUIRED=$((FAILED_REQUIRED + 1))
+        echo "    ❌ Required test suite failed" >&2
+      fi
+    }
+
+    echo "$RESULT" >> "$RESULTS_FILE"
+  done <<< "$TEST_SUITES"
+
+  # If any required test suite failed completely, gate fails immediately
+  if [ $FAILED_REQUIRED -gt 0 ]; then
+    echo ""
+    echo "❌ Gate FAILED: $FAILED_REQUIRED required test suite(s) failed completely" >&2
+    generate_iteration_context "$TASK_ID" "0.0" "$(get_mode_threshold "$MODE")" "$RESULTS_FILE"
+    return 1
+  fi
+
+  # Calculate aggregate pass rate
+  local TOTAL_PASS_RATE
+  TOTAL_PASS_RATE=$(calculate_aggregate_pass_rate "$RESULTS_FILE")
+
+  # Store results
+  store_test_results "$TASK_ID" "$TOTAL_PASS_RATE" "$RESULTS_FILE"
+
+  # Get threshold for mode
+  local TEST_THRESHOLD
+  TEST_THRESHOLD=$(get_mode_threshold "$MODE")
+
+  echo ""
+  echo "Test Results Summary:"
+  echo "  Aggregate Pass Rate: $TOTAL_PASS_RATE"
+  echo "  Required Threshold: $TEST_THRESHOLD ($MODE mode)"
+  echo ""
+
+  # Check threshold
+  if (( $(echo "$TOTAL_PASS_RATE >= $TEST_THRESHOLD" | bc -l) )); then
+    echo "✅ Gate PASSED: Test-driven validation successful"
+    return 0
+  else
+    echo "❌ Gate FAILED: Pass rate below threshold"
+    echo "   Gap: $(echo "$TEST_THRESHOLD - $TOTAL_PASS_RATE" | bc -l)"
+    generate_iteration_context "$TASK_ID" "$TOTAL_PASS_RATE" "$TEST_THRESHOLD" "$RESULTS_FILE"
+    return 1
+  fi
+}
+
+##############################################################################
+# Confidence-Based Gate Check (Original Implementation)
+##############################################################################
+
+gate_check_confidence() {
+  local TASK_ID="$1"
+  local AGENTS="$2"
+  local THRESHOLD="$3"
+  local MIN_QUORUM="$4"
+
+  # Deprecation warning
+  echo "⚠️  Using legacy confidence-based gate check" >&2
+  echo "   Consider migrating to test-driven validation" >&2
+  echo ""
+
+  # Use Redis Coordination skill to collect confidence scores
+  local REDIS_COORD_SKILL="$PROJECT_ROOT/.claude/skills/cfn-redis-coordination"
+
+  echo "Gate Check Configuration:"
+  echo "  Task ID: $TASK_ID"
+  echo "  Agent IDs: $AGENTS"
+  echo "  Min Quorum: $MIN_QUORUM"
+  echo ""
+
+  # Collect Loop 3 confidence scores
+  local CONSENSUS
+  CONSENSUS=$("$REDIS_COORD_SKILL/invoke-waiting-mode.sh" collect \
+    --task-id "$TASK_ID" \
+    --agent-ids "$AGENTS" \
+    --min-quorum "$MIN_QUORUM" 2>/dev/null) || {
+    echo "❌ Error: Failed to collect Loop 3 confidence scores" >&2
+    echo "   Agent IDs: $AGENTS" >&2
+    return 1
+  }
+
+  # Validate consensus is a valid number
+  if ! [[ "$CONSENSUS" =~ ^[0-9]+\.?[0-9]*$ ]]; then
+    echo "⚠️  WARNING: Invalid consensus value: $CONSENSUS (expected numeric)" >&2
+    echo "   Defaulting to 0.0" >&2
+    CONSENSUS="0.0"
+  fi
+
+  echo "Loop 3 Gate Check:"
+  echo "  Consensus: $CONSENSUS"
+  echo "  Threshold: $THRESHOLD"
+  echo "  Required: >= $THRESHOLD"
+  echo ""
+
+  # Compare consensus to gate threshold
+  if (( $(echo "$CONSENSUS >= $THRESHOLD" | bc -l) )); then
+    echo "✅ Gate PASSED - Loop 3 self-validation successful"
+    return 0
+  else
+    echo "❌ Gate FAILED - Loop 3 needs improvement"
+    echo "   Gap: $(echo "$THRESHOLD - $CONSENSUS" | bc -l)"
+    return 1
+  fi
+}
+
+##############################################################################
+# Main Execution - Strategy Selection
+##############################################################################
+
+echo "========================================="
+echo "CFN Loop Gate Check v2.0"
+echo "Strategy: $STRATEGY"
+echo "========================================="
+echo ""
+
+case "$STRATEGY" in
+  test-driven)
+    if [ -z "$SUCCESS_CRITERIA" ]; then
+      echo "❌ Error: test-driven strategy requires --success-criteria" >&2
+      exit 1
+    fi
+    gate_check_test_driven "$TASK_ID" "$MODE" "$SUCCESS_CRITERIA"
+    ;;
+
+  confidence)
+    gate_check_confidence "$TASK_ID" "$AGENTS" "$THRESHOLD" "$MIN_QUORUM"
+    ;;
+
+  auto)
+    # Hybrid mode: prefer test-driven if criteria exists, fallback to confidence
+    if has_success_criteria; then
+      echo "Auto-detected: Using test-driven validation" >&2
+      echo ""
+      gate_check_test_driven "$TASK_ID" "$MODE" "$SUCCESS_CRITERIA"
+    else
+      echo "Auto-detected: Using confidence-based validation (no success criteria)" >&2
+      echo ""
+      gate_check_confidence "$TASK_ID" "$AGENTS" "$THRESHOLD" "$MIN_QUORUM"
+    fi
+    ;;
+
+  *)
+    echo "❌ Error: Invalid strategy '$STRATEGY'" >&2
+    echo "   Valid options: test-driven, confidence, auto" >&2
+    exit 1
+    ;;
+esac
+
+exit $?