claude-flow-novice 2.15.4 → 2.15.6

package/claude-assets/skills/cfn-loop-orchestration/orchestrate.sh

@@ -16,7 +16,7 @@
 #                    [--max-iterations <n>] \
 #                    [--epic-context <json>] \
 #                    [--phase-context <json>] \
-#                    [--success-criteria <json>]
+#                    [--success-criteria <enabled>]  # Flag: criteria stored in Redis by coordinator
 ##############################################################################
 
 set -euo pipefail
@@ -61,6 +61,21 @@ mkdir -p "$CFN_TELEMETRY_DIR"
 # shellcheck source=./security_utils.sh
 source "$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)/security_utils.sh"
 
+##############################################################################
+# Input Sanitization (Security)
+##############################################################################
+sanitize_input() {
+  local input="$1"
+  local max_length="${2:-256}"  # Default max length 256 chars
+
+  # Truncate to max length
+  input="${input:0:$max_length}"
+
+  # Remove dangerous characters (only allow alphanumeric, dash, underscore, dot, comma, colon, space, forward slash)
+  # This covers task IDs, agent types, file paths, and JSON-like structures
+  echo "$input" | sed 's/[^a-zA-Z0-9._:, /-]//g'
+}
+
 HELPERS_DIR="$SCRIPT_DIR/helpers"
 REDIS_COORD_SKILL="$PROJECT_ROOT/.claude/skills/cfn-redis-coordination"
 
@@ -142,6 +157,10 @@ while [[ $# -gt 0 ]]; do
         echo "Error: --loop3-agents requires a value"
         exit 1
       fi
+      if [[ -z "$2" ]]; then
+        echo "Error: --loop3-agents value cannot be empty"
+        exit 1
+      fi
       validate_agent_list "$2" || { echo "Invalid Loop 3 agent list"; exit 1; }
       LOOP3_AGENTS="$2"
       shift 2
@@ -151,6 +170,10 @@ while [[ $# -gt 0 ]]; do
         echo "Error: --loop2-agents requires a value"
         exit 1
       fi
+      if [[ -z "$2" ]]; then
+        echo "Error: --loop2-agents value cannot be empty"
+        exit 1
+      fi
       validate_agent_list "$2" || { echo "Invalid Loop 2 agent list"; exit 1; }
       LOOP2_AGENTS="$2"
       shift 2
@@ -160,6 +183,10 @@ while [[ $# -gt 0 ]]; do
         echo "Error: --product-owner requires a value"
         exit 1
       fi
+      if [[ -z "$2" ]]; then
+        echo "Error: --product-owner value cannot be empty"
+        exit 1
+      fi
       PRODUCT_OWNER=$(sanitize_input "$2") || { echo "Invalid product owner"; exit 1; }
       shift 2
       ;;
@@ -236,7 +263,8 @@ while [[ $# -gt 0 ]]; do
         echo "Error: --success-criteria requires a value"
         exit 1
       fi
-      validate_json_context "$2" || { echo "Invalid success criteria JSON"; exit 1; }
+      # Store flag value - coordinator will write to Redis before spawning orchestrator
+      # Orchestrator validates the key exists in Redis during pre-flight
       SUCCESS_CRITERIA="$2"
       shift 2
       ;;
@@ -281,7 +309,7 @@ while [[ $# -gt 0 ]]; do
       echo "  --min-quorum-loop2 <n>      Loop 2 quorum threshold (default: 0.66)"
       echo "  --epic-context <json>       Epic context JSON"
       echo "  --phase-context <json>      Phase context JSON"
-      echo "  --success-criteria <json>   Success criteria JSON"
+      echo "  --success-criteria <value>  Flag to enable success criteria (stored in Redis by coordinator)"
       echo "  --expected-files <files>    Comma-separated expected deliverables"
       echo "  --phase-id <id>             Phase identifier for timeout calculation"
       exit 1
@@ -342,6 +370,81 @@ echo "Timeout: ${TIMEOUT}s"
 echo "=============================================="
 echo ""
 
+##############################################################################
+# Pre-Flight Validation
+##############################################################################
+
+echo "Running pre-flight checks..."
+
+# 1. Validate Success Criteria in Redis (if flag provided)
+if [ -n "$SUCCESS_CRITERIA" ]; then
+  CRITERIA_VALUE=$("$REDIS_COORD_SKILL/get-context.sh" \
+    --task-id "$TASK_ID" \
+    --key "success-criteria" \
+    --namespace "swarm" 2>/dev/null || echo "")
+
+  if [ -z "$CRITERIA_VALUE" ]; then
+    echo "❌ Pre-flight failed: --success-criteria flag set but not found in Redis" >&2
+    echo "   Coordinator must store criteria before spawning orchestrator" >&2
+    exit 1
+  fi
+
+  # Validate JSON syntax
+  if ! echo "$CRITERIA_VALUE" | jq empty 2>/dev/null; then
+    echo "❌ Pre-flight failed: Success criteria in Redis contains invalid JSON" >&2
+    exit 1
+  fi
+
+  echo "✅ Success criteria validated in Redis"
+fi
+
+# 2. Validate Redis Connectivity (required for CLI mode)
+if ! redis-cli -h "${REDIS_HOST:-localhost}" -p "${REDIS_PORT:-6379}" ping &>/dev/null; then
+  echo "❌ Pre-flight failed: Redis not available at ${REDIS_HOST:-localhost}:${REDIS_PORT:-6379}" >&2
+  echo "   CLI mode requires Redis for coordination" >&2
+  exit 1
+fi
+echo "✅ Redis connectivity validated"
+
+# 3. Validate Agent Types Exist
+MISSING_AGENTS=""
+for agent_type in $(echo "$LOOP3_AGENTS,$LOOP2_AGENTS" | tr ',' '\n' | sort -u); do
+  if ! find "$PROJECT_ROOT/.claude/agents" -name "*${agent_type}*.md" 2>/dev/null | grep -q .; then
+    MISSING_AGENTS="${MISSING_AGENTS}${agent_type}, "
+  fi
+done
+
+if [ -n "$MISSING_AGENTS" ]; then
+  echo "⚠️  Warning: Agent types not found in .claude/agents/: ${MISSING_AGENTS%, }" >&2
+  echo "   Continuing anyway (agent-spawn will handle missing agents)" >&2
+fi
+
+# 4. Validate Helper Scripts Exist
+REQUIRED_HELPERS=(
+  "$HELPERS_DIR/gate-check.sh"
+  "$HELPERS_DIR/consensus.sh"
+  "$HELPERS_DIR/iteration-manager.sh"
+  "$HELPERS_DIR/timeout-calculator.sh"
+)
+
+for helper in "${REQUIRED_HELPERS[@]}"; do
+  if [ ! -x "$helper" ]; then
+    echo "❌ Pre-flight failed: Required helper script missing or not executable: $helper" >&2
+    exit 1
+  fi
+done
+echo "✅ Helper scripts validated"
+
+# 5. Validate Product Owner Decision Skill
+if [ ! -x "$PROJECT_ROOT/.claude/skills/cfn-product-owner-decision/execute-decision.sh" ]; then
+  echo "❌ Pre-flight failed: Product owner decision script not found or not executable" >&2
+  exit 1
+fi
+echo "✅ Product owner decision script validated"
+
+echo "✅ All pre-flight checks passed"
+echo ""
+
 ##############################################################################
 # Helper Functions
 ##############################################################################
@@ -369,14 +472,21 @@ function store_context() {
     echo "Stored phase context"
   fi
 
-  # Store success criteria if provided using Redis coordination primitive
+  # NOTE: Success criteria is now stored by coordinator BEFORE spawning orchestrator
+  # Orchestrator only validates that it exists in Redis during pre-flight
+  # This comment explains the flow for future maintainers
   if [ -n "$SUCCESS_CRITERIA" ]; then
-    "$REDIS_COORD_SKILL/store-context.sh" \
+    # Verify criteria exists in Redis (should have been stored by coordinator)
+    CRITERIA_VALUE=$("$REDIS_COORD_SKILL/get-context.sh" \
       --task-id "$task_id" \
       --key "success-criteria" \
-      --value "$SUCCESS_CRITERIA" \
-      --namespace "swarm" >/dev/null
-    echo "Stored success criteria"
+      --namespace "swarm" 2>/dev/null || echo "")
+
+    if [ -n "$CRITERIA_VALUE" ]; then
+      echo "✅ Success criteria loaded from Redis (stored by coordinator)"
+    else
+      echo "⚠️  Success criteria flag set but not found in Redis" >&2
+    fi
   fi
 
   echo ""
@@ -553,22 +663,43 @@ function spawn_loop3_agents() {
     safe_agent_id=$(sanitize_input "$UNIQUE_AGENT_ID") || continue
 
     # Dual-mode agent spawning: Docker or CLI
-    # Docker mode: CFN_DOCKER_MODE=true or Docker socket available
-    # CLI mode: Default (uses npx)
-    if [[ "${CFN_DOCKER_MODE:-false}" == "true" ]] || [[ -S /var/run/docker.sock ]]; then
+    # Mode Selection Priority:
+    # 1. Explicit CFN_DOCKER_MODE='true'/'false' (highest priority - user override)
+    # 2. Automatic Docker socket detection (if CFN_DOCKER_MODE unset)
+    # 3. Default CLI mode (fallback if no Docker socket)
+    #
+    # BUG FIX: Respect CFN_DOCKER_MODE='false' even when Docker socket exists
+    SPAWN_MODE="cli"  # Default
+    SPAWN_REASON=""
+
+    if [[ "${CFN_DOCKER_MODE:-}" == "true" ]]; then
+        SPAWN_MODE="docker"
+        SPAWN_REASON="explicit CFN_DOCKER_MODE=true"
+    elif [[ "${CFN_DOCKER_MODE:-}" == "false" ]]; then
+        SPAWN_MODE="cli"
+        SPAWN_REASON="explicit CFN_DOCKER_MODE=false (overrides Docker socket detection)"
+    elif [[ -S /var/run/docker.sock ]]; then
+        SPAWN_MODE="docker"
+        SPAWN_REASON="automatic Docker socket detection"
+    else
+        SPAWN_MODE="cli"
+        SPAWN_REASON="default (no Docker socket)"
+    fi
+
+    if [[ "$SPAWN_MODE" == "docker" ]]; then
         # Docker-based spawning (prevents WebAssembly OOM)
-        echo "  → Docker mode: spawning via container" >&2
+        echo "  → Docker mode: ${SPAWN_REASON}" >&2
 
         # SECURITY FIX: Sanitize Docker environment variables to prevent command injection
-        CFN_DOCKER_IMAGE_SAFE=$(sanitize_docker_var "${CFN_DOCKER_IMAGE:-claude-flow-novice:agent}") || {
+        CFN_DOCKER_IMAGE_SAFE=$(sanitize_input "${CFN_DOCKER_IMAGE:-claude-flow-novice:agent}") || {
           echo "❌ Invalid CFN_DOCKER_IMAGE" >&2
           exit 1
         }
-        CFN_DOCKER_NETWORK_SAFE=$(sanitize_docker_var "${CFN_DOCKER_NETWORK:-mcp-network}") || {
+        CFN_DOCKER_NETWORK_SAFE=$(sanitize_input "${CFN_DOCKER_NETWORK:-mcp-network}") || {
           echo "❌ Invalid CFN_DOCKER_NETWORK" >&2
           exit 1
         }
-        CFN_MEMORY_LIMIT_SAFE=$(sanitize_docker_var "${CFN_MEMORY_LIMIT:-2g}") || {
+        CFN_MEMORY_LIMIT_SAFE=$(sanitize_input "${CFN_MEMORY_LIMIT:-2g}") || {
           echo "❌ Invalid CFN_MEMORY_LIMIT" >&2
           exit 1
         }
@@ -619,7 +750,7 @@ function spawn_loop3_agents() {
         AGENT_PID=$!
     else
         # CLI-based spawning (traditional approach)
-        echo "  → CLI mode: spawning via npx" >&2
+        echo "  → CLI mode: ${SPAWN_REASON}" >&2
 
         if command -v execute_instrumented >/dev/null 2>&1; then
             execute_instrumented "npx" "$CFN_VALIDATION_TIMEOUT" "$CFN_MEMORY_LIMIT" \