1 files changed, 62 insertions, 5 deletions

diff --git a/internal/executor/executor.go b/internal/executor/executor.go
index 219a40b..1f2c27d 100644
--- a/internal/executor/executor.go
+++ b/internal/executor/executor.go
@@ -30,6 +30,7 @@ type Store interface {
 	AppendTaskInteraction(taskID string, interaction task.Interaction) error
 	UpdateTaskAgent(id string, agent task.AgentConfig) error
 	UpdateExecutionChangestats(execID string, stats *task.Changestats) error
+	RecordAgentEvent(e storage.AgentEvent) error
 }
 
 // LogPather is an optional interface runners can implement to provide the log
@@ -273,16 +274,32 @@ func (p *Pool) handleRunResult(ctx context.Context, t *task.Task, exec *storage.
 		if isRateLimitError(err) || isQuotaExhausted(err) {
 			p.mu.Lock()
 			retryAfter := parseRetryAfter(err.Error())
-			if retryAfter == 0 {
-				if isQuotaExhausted(err) {
+			reason := "transient"
+			if isQuotaExhausted(err) {
+				reason = "quota"
+				if retryAfter == 0 {
 					retryAfter = 5 * time.Hour
-				} else {
-					retryAfter = 1 * time.Minute
 				}
+			} else if retryAfter == 0 {
+				retryAfter = 1 * time.Minute
 			}
-			p.rateLimited[agentType] = time.Now().Add(retryAfter)
+			until := time.Now().Add(retryAfter)
+			p.rateLimited[agentType] = until
 			p.logger.Info("agent rate limited", "agent", agentType, "retryAfter", retryAfter, "quotaExhausted", isQuotaExhausted(err))
 			p.mu.Unlock()
+			go func() {
+				ev := storage.AgentEvent{
+					ID:        uuid.New().String(),
+					Agent:     agentType,
+					Event:     "rate_limited",
+					Timestamp: time.Now(),
+					Until:     &until,
+					Reason:    reason,
+				}
+				if recErr := p.store.RecordAgentEvent(ev); recErr != nil {
+					p.logger.Warn("failed to record agent event", "error", recErr)
+				}
+			}()
 		}
 
 		var blockedErr *BlockedError
@@ -382,6 +399,34 @@ func (p *Pool) ActiveCount() int {
 	return p.active
 }
 
+// AgentStatusInfo holds the current state of a single agent.
+type AgentStatusInfo struct {
+	Agent       string     `json:"agent"`
+	ActiveTasks int        `json:"active_tasks"`
+	RateLimited bool       `json:"rate_limited"`
+	Until       *time.Time `json:"until,omitempty"`
+}
+
+// AgentStatuses returns the current status of all registered agents.
+func (p *Pool) AgentStatuses() []AgentStatusInfo {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	now := time.Now()
+	var out []AgentStatusInfo
+	for agent := range p.runners {
+		info := AgentStatusInfo{
+			Agent:       agent,
+			ActiveTasks: p.activePerAgent[agent],
+		}
+		if deadline, ok := p.rateLimited[agent]; ok && now.Before(deadline) {
+			info.RateLimited = true
+			info.Until = &deadline
+		}
+		out = append(out, info)
+	}
+	return out
+}
+
 // pickAgent selects the best agent from the given SystemStatus using explicit
 // load balancing: prefer the available (non-rate-limited) agent with the fewest
 // active tasks. If all agents are rate-limited, fall back to fewest active.
@@ -423,6 +468,18 @@ func (p *Pool) execute(ctx context.Context, t *task.Task) {
 		activeTasks[agent] = p.activePerAgent[agent]
 		if deadline, ok := p.rateLimited[agent]; ok && now.After(deadline) {
 			delete(p.rateLimited, agent)
+			agentName := agent
+			go func() {
+				ev := storage.AgentEvent{
+					ID:        uuid.New().String(),
+					Agent:     agentName,
+					Event:     "available",
+					Timestamp: time.Now(),
+				}
+				if recErr := p.store.RecordAgentEvent(ev); recErr != nil {
+					p.logger.Warn("failed to record agent available event", "error", recErr)
+				}
+			}()
 		}
 		rateLimited[agent] = now.Before(p.rateLimited[agent])
 	}