feat: output channel redesign - per-channel output gates, LLM chain events, SDKConfig

- Output channels generate per-channel tools: output_send__{name} (type=output) + output_send__{name}_help
- content is JSON string transparently passed to plugin handler for routing
- EventAgentLLMChain: full LLM response forwarded after each turn for webui/logs
- sdk.New refactored to SDKConfig struct (no more 13 positional args)
- RegisterOutputChannel adds desc param for JSON format documentation
- channelDevice simplified (no Tools method), desc field added
- Child agent permission updated for output_send__ prefix
- System prompt: output gates, multi-call, long messages split
- WebUI: subscribes to EventAgentLLMChain in SSE, no output channel
- Tests updated for new naming convention

internal/agent/core/plugin_health.go

@@ -0,0 +1,128 @@
+package core
+
+import (
+	"log"
+	"sync"
+	"time"
+)
+
+const (
+	maxPluginCrashes  = 3
+	crashWindow       = 5 * time.Minute
+	reloadCooldown    = 30 * time.Second
+)
+
+type pluginHealthTracker struct {
+	mu      sync.Mutex
+	records map[string]*pluginHealthRecord
+}
+
+type pluginHealthRecord struct {
+	CrashCount int
+	FirstCrash time.Time
+	LastCrash  time.Time
+	Unhealthy  bool
+	LastReload time.Time
+}
+
+func newPluginHealthTracker() *pluginHealthTracker {
+	return &pluginHealthTracker{
+		records: make(map[string]*pluginHealthRecord),
+	}
+}
+
+// recordCrash 记录一次崩溃，返回 true 表示需要触发重载
+func (t *pluginHealthTracker) recordCrash(plugin string) bool {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+
+	now := time.Now()
+	r, ok := t.records[plugin]
+	if !ok {
+		r = &pluginHealthRecord{}
+		t.records[plugin] = r
+	}
+
+	if now.Sub(r.LastCrash) > crashWindow {
+		r.CrashCount = 0
+		r.FirstCrash = now
+	}
+
+	r.CrashCount++
+	r.LastCrash = now
+
+	if r.CrashCount >= maxPluginCrashes {
+		r.Unhealthy = true
+		log.Printf("[plugin] %s: %d crashes within %v, marking unhealthy", plugin, r.CrashCount, crashWindow)
+		return true
+	}
+
+	log.Printf("[plugin] %s: crash #%d", plugin, r.CrashCount)
+	return false
+}
+
+// isHealthy 检查插件是否健康；冷却期后自动恢复
+func (t *pluginHealthTracker) isHealthy(plugin string) bool {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+
+	r, ok := t.records[plugin]
+	if !ok {
+		return true
+	}
+	if !r.Unhealthy {
+		return true
+	}
+	if time.Since(r.LastReload) > reloadCooldown {
+		r.Unhealthy = false
+		r.CrashCount = 0
+		log.Printf("[plugin] %s: cooldown passed, restored to healthy", plugin)
+		return true
+	}
+	return false
+}
+
+// markReloaded 标记插件已重载
+func (t *pluginHealthTracker) markReloaded(plugin string) {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+
+	r, ok := t.records[plugin]
+	if ok {
+		r.Unhealthy = false
+		r.CrashCount = 0
+		r.LastReload = time.Now()
+	}
+}
+
+// pendingReloads 返回已过冷却期、需要重载的插件列表
+func (t *pluginHealthTracker) pendingReloads() []string {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+
+	var result []string
+	now := time.Now()
+	for name, r := range t.records {
+		if !r.Unhealthy {
+			continue
+		}
+		if now.Sub(r.LastReload) > reloadCooldown {
+			result = append(result, name)
+		}
+	}
+	return result
+}
+
+// unhealthyPlugins 返回当前所有不健康的插件名
+func (t *pluginHealthTracker) unhealthyPlugins() []string {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+
+	var result []string
+	for name, r := range t.records {
+		if r.Unhealthy {
+			result = append(result, name)
+		}
+	}
+	return result
+}