claude-flow-novice 2.14.3 → 2.14.4

package/.claude/skills/cfn-redis-coordination/SHUTDOWN_HANDLING.md

@@ -1,164 +0,0 @@
-# Shutdown Handling for Waiting Agents
-
-## Overview
-
-Agents in waiting mode can now listen for shutdown signals and exit gracefully with proper cleanup. This enables coordinators to terminate long-running tasks without leaving orphaned agents in the system.
-
-## Implementation Details
-
-### Multi-Key Monitoring
-
-Agents in waiting mode now monitor **two** Redis keys simultaneously:
-1. **Wake Queue**: `swarm:{task-id}:{agent-id}:wake-queue` (priority queue for wake signals)
-2. **Shutdown Key**: `swarm:{task-id}:shutdown` (broadcast shutdown signal)
-
-### Polling Strategy
-
-Since Redis `BZPOPMIN` doesn't support multiple keys, we implement a polling loop:
-
-```bash
-while true; do
-    # Check shutdown first (highest priority)
-    SHUTDOWN_CHECK=$(redis-cli LPOP "swarm:${TASK_ID}:shutdown")
-
-    if shutdown_detected; then
-        exit 130  # SIGINT code for graceful shutdown
-    fi
-
-    # Block on wake queue with 1-second timeout
-    WAKE_RESULT=$(redis-cli BZPOPMIN "$WAKE_QUEUE" 1)
-
-    if wake_received; then
-        process_wake_signal
-        break
-    fi
-done
-```
-
-### Exit Code Convention
-
-- **Exit Code 130**: Graceful shutdown via shutdown signal (SIGINT equivalent)
-- **Exit Code 0**: Normal completion after wake signal
-
-## Usage
-
-### Coordinator Broadcasts Shutdown
-
-```bash
-./invoke-waiting-mode.sh shutdown \
-  --task-id "auth-system" \
-  --reason "task_complete"
-```
-
-**Broadcast Behavior:**
-- Shutdown signal is added to `swarm:{task-id}:shutdown` key
-- All agents monitoring this task will receive the signal on their next poll cycle (max 1 second latency)
-- Multiple agents can be shutdown with a single broadcast
-
-### Agent Enters Waiting Mode
-
-```bash
-./invoke-waiting-mode.sh enter \
-  --task-id "auth-system" \
-  --agent-id "coder-1" \
-  --context "iteration-1"
-```
-
-**Agent will:**
-1. Publish ready status to Redis
-2. Enter polling loop monitoring both wake queue and shutdown key
-3. Check shutdown key first (highest priority)
-4. Block on wake queue with 1-second timeout
-5. Exit with code 130 if shutdown signal received
-6. Continue normally if wake signal received
-
-## Priority Behavior
-
-**Shutdown signals have HIGHEST priority:**
-- Checked **before** wake queue on every poll cycle
-- Even if wake signals are in the queue, shutdown is processed first
-- Ensures clean task termination
-
-## Performance Characteristics
-
-- **Latency**: Max 1 second to detect shutdown (BZPOPMIN timeout)
-- **Cost**: Still zero tokens during waiting (BZPOPMIN blocks, LPOP is instant)
-- **Overhead**: Minimal - one LPOP per second for shutdown check
-
-## Test Results
-
-```bash
-bash .claude/skills/redis-coordination/test-shutdown.sh
-```
-
-**Test Coverage:**
-1. ✅ Agent enters waiting mode successfully
-2. ✅ Agent receives shutdown signal and exits with code 130
-3. ✅ Shutdown signal is broadcasted properly
-4. ✅ Agent terminates gracefully within 1-2 seconds
-
-## Integration Examples
-
-### CFN Loop Coordinator
-
-```bash
-# After consensus reached or max iterations
-./invoke-waiting-mode.sh shutdown \
-  --task-id "$TASK_ID" \
-  --reason "cfn_loop_complete"
-
-echo "All agents notified of task completion"
-```
-
-### Emergency Shutdown
-
-```bash
-# Abort all agents in a task
-./invoke-waiting-mode.sh shutdown \
-  --task-id "$TASK_ID" \
-  --reason "emergency_abort"
-```
-
-### Timeout Management
-
-```bash
-# Coordinator timeout
-timeout 300 ./my-coordination-script.sh || {
-  echo "Coordinator timeout - sending shutdown to all agents"
-  ./invoke-waiting-mode.sh shutdown \
-    --task-id "$TASK_ID" \
-    --reason "coordinator_timeout"
-}
-```
-
-## File Changes
-
-**Modified:**
-- `/mnt/c/Users/masha/Documents/claude-flow-novice/.claude/skills/redis-coordination/invoke-waiting-mode.sh`
-
-**Added:**
-- `shutdown` command for broadcasting shutdown signals
-- Multi-key monitoring in `enter` command
-- Polling loop with shutdown priority
-- Exit code 130 convention for graceful shutdown
-
-**Test Suite:**
-- `/mnt/c/Users/masha/Documents/claude-flow-novice/.claude/skills/redis-coordination/test-shutdown.sh`
-
-## Benefits
-
-1. **Graceful Cleanup**: Agents exit cleanly instead of hanging indefinitely
-2. **Broadcast Efficiency**: Single shutdown command affects all agents
-3. **Priority Handling**: Shutdown always takes precedence over wake signals
-4. **Zero Token Cost**: Still maintains efficient blocking while waiting
-5. **Fast Response**: Max 1 second latency to detect shutdown
-
-## Future Enhancements
-
-Potential improvements for future iterations:
-
-1. **Agent-Specific Shutdown**: Target individual agents instead of broadcast
-2. **Shutdown Acknowledgement**: Agents report shutdown receipt before exiting
-3. **Lua Script Optimization**: Atomic shutdown check + wake queue pop
-4. **Timeout Configuration**: Configurable poll timeout instead of hardcoded 1 second
-5. **Shutdown Reason Logging**: Persist shutdown events to Redis for audit trail